Module:form of: Difference between revisions

From Linguifex
Jump to navigation Jump to search
No edit summary
No edit summary
 
(6 intermediate revisions by the same user not shown)
Line 1: Line 1:
local export = {}
export.force_cat = false -- for testing; set to true to display categories even on non-mainspace pages
local m_links = require("Module:links")
local m_links = require("Module:links")
local m_string_utils = require("Module:string utilities")
local m_table = require("Module:table")
local m_table = require("Module:table")
local m_pos = mw.loadData("Module:form of/pos")
local parse_utilities_module = "Module:parse utilities"
local m_functions = require("Module:form of/functions")
local labels_module = "Module:labels"
local utilities_module = "Module:utilities"
export.form_of_pos_module = "Module:form of/pos"
export.form_of_functions_module = "Module:form of/functions"
export.form_of_cats_module = "Module:form of/cats"
export.form_of_lang_data_module_prefix = "Module:form of/lang-data/"
export.form_of_data_module = "Module:form of/data"
export.form_of_data2_module = "Module:form of/data2"


local ulen = mw.ustring.len
local ulen = m_string_utils.len
local rsubn = mw.ustring.gsub
local rsubn = m_string_utils.gsub
local rmatch = mw.ustring.match
local rmatch = m_string_utils.match
local rsplit = mw.text.split
local rsplit = m_string_utils.split


local export = {}
export.TAG_TYPE = 1
export.GLOSSARY = 2
export.SHORTCUTS = 3
export.WIKIDATA = 4


--[=[
export.APPENDIX = true
export.WP = false
export.WIKT = 0


This module implements the underlying processing of {{form of}},
--[==[
{{inflection of}} and specific variants such as {{past participle of}}
Set listing the languages with lang-specific tags. If a language isn't listed here, the tags for that language won't be
and {{alternative spelling of}}. Most of the logic in this file is to
recognized.
handle tags in {{inflection of}}. Other related files:
]==]
export.langs_with_lang_specific_tags = {
["en"] = true,
["got"] = true,
["lt"] = true,
["lv"] = true,
["nl"] = true,
["pi"] = true,
["sw"] = true,
["ttj"] = true,
}


* [[Module:form of/templates]] contains the majority of the logic that
--[==[ intro:
  implements the templates themselves.
* [[Module:form of/data]] is a data-only file containing information on
  the more common inflection tags, listing the tags, their shortcuts,
  the category they belong to (tense-aspect, case, gender, voice-valence,
  etc.), the appropriate glossary link and the wikidata ID.
* [[Module:form of/data2]] is a data-only file containing information on
  the less common inflection tags, in the same format as
  [[Module:form of/data]].
* [[Module:form of/cats]] is a data-only file listing the
  language-specific categories that are added when the appropriate
  combinations of tags are seen for a given language.
* [[Module:form of/pos]] is a data-only file listing the recognized
  parts of speech and their abbreviations, used for categorization.
  FIXME: This should be unified with the parts of speech listed in
  [[Module:links]].
* [[Module:form of/functions]] contains functions for use with
  [[Module:form of/data]] and [[Module:form of/cats]]. They are
  contained in this module because data-only modules can't contain
  code. The functions in this file are of two types:


  (1) Display handlers allow for customization of the display of
This module implements the underlying processing of {{tl|form of}}, {{tl|inflection of}} and specific variants such as
      multipart tags (see below). Currently there is only one
{{tl|past participle of}} and {{tl|alternative spelling of}}. Most of the logic in this file is to handle tags in
  such handler, for handling multipart person tags such as
{{tl|inflection of}}. Other related files:
  '1//2//3'.
  (2) Cat functions allow for more complex categorization logic,
      and are referred to by name in [[Module:form of/cats]].
  Currently no such functions exist.


The following terminology is used in conjunction with {{inflection of}}:
* [[Module:form of/templates]] contains the majority of the logic that implements the templates themselves.
* [[Module:form of/data]] is a data-only file containing information on the more common inflection tags, listing the
  tags, their shortcuts, the category they belong to (tense-aspect, case, gender, voice-valence, etc.), the appropriate
  glossary link and the wikidata ID.
* [[Module:form of/data2]] is a data-only file containing information on the less common inflection tags, in the same
  format as [[Module:form of/data]].
* [[Module:form of/lang-data/LANGCODE]] is a data-only file containing information on the language-specific inflection
  tags for the language with code LANGCODE, in the same format as [[Module:form of/data]]. Language-specific tags
  override general tags.
* [[Module:form of/cats]] is a data-only file listing the language-specific categories that are added when the
  appropriate combinations of tags are seen for a given language.
* [[Module:form of/pos]] is a data-only file listing the recognized parts of speech and their abbreviations, used for
  categorization. FIXME: This should be unified with the parts of speech listed in [[Module:links]].
* [[Module:form of/functions]] contains functions for use with [[Module:form of/data]] and [[Module:form of/cats]].
  They are contained in this module because data-only modules can't contain code. The functions in this file are of two
  types:
*# Display handlers allow for customization of the display of multipart tags (see below). Currently there is only
  one such handler, for handling multipart person tags such as `1//2//3`.
*# Cat functions allow for more complex categorization logic, and are referred to by name in [[Module:form of/cats]].
  Currently no such functions exist.


* A TAG is a single grammatical item, as specified in a single numbered
The following terminology is used in conjunction with {{tl|inflection of}}:
  parameter of {{inflection of}}. Examples are 'masculine', 'nominative',
  or 'first-person'. Tags may be abbreviated, e.g. 'm' for 'masculine',
  'nom' for 'nominative', or '1' for 'first-person'. Such abbreviations
  are called SHORTCUTS, and some tags have multiple equivalent shortcuts
  (e.g. 'p' or 'pl' for 'plural'). The full, non-abbreviated form of
  a tag is called its CANONICAL FORM.
* The DISPLAY FORM of a tag is the way it's displayed to the user. Usually
  the displayed text of the tag is the same as its canonical form, and it
  normally functions as a link to a glossary entry explaining the tag.
  Usually the link is to an entry in [[Appendix:Glossary]], but sometimes
  the tag is linked to an individual dictionary entry or to a Wikipedia
  entry. Occasionally, the display text differs from the canonical form of
  the tag. An example is the tag 'comparative case', which has the display
  text read as simply 'comparative'. Normally, tags referring to cases don't
  have the word "case" in them, but in this case the tag 'comparative' was
  already used as a shortcut for the tag 'comparative degree', so the tag was
  named 'comparative case' to avoid clashing. A similar situation occurs
  with 'adverbial case' vs. the grammar tag 'adverbial' (as in 'adverbial
  participle').
* A TAG SET is an ordered list of tags, which together express a single
  inflection, for example, '1|s|pres|ind', which can be expanded to
  canonical-form tags as 'first-person|singular|present|indicative'.
  Multiple tag sets can be specified in a single call to {{inflection of}}
  by separating the individual tag sets with a semicolon, e.g.
  '1|s|pres|ind|;|2|s|imp', which specifies two tag sets, '1|s|pres|ind'
  as above and '2|s|imp' (in canonical form,
  'second-person|singular|imperative').
* A MULTIPART TAG is a tag that embeds multiple tags within it, such as
  'f//n' or 'nom//acc//voc'. These are used in the case of [[syncretism]],
  when the same form applies to multiple inflections. Examples are the
  Spanish present subjunctive, where the first-person and third-person
  singular have the same form (e.g. [[siga]] from [[seguir]] "to follow"),
  or Latin third-declension adjectives, where the dative and ablative
  plural of all genders have the same form (e.g. [[omnibus]] from [[omnis]]
  "all"). These would be expressed respectively as '1//3|s|pres|sub'
  and 'dat//abl|m//f//n|p', where the use of the multipart tag compactly
  encodes the syncretism and avoids the need to individually list out
  all of the inflections. Multipart tags currently display as a list
  separated by "and", ''dative and ablative'' or
  ''masculine, feminine and neuter'' where each individual word is linked
  appropriately. As a special case, multipart tags involving persons display
  specially; for example, the multipart tag ''1//2//3'' displays as
  ''first-, second- and third-person'', with the word "person" occurring
  only once.
* A TWO-LEVEL MULTIPART TAG is a special type of multipart tag that
  joins two or more tag sets instead of joining individual tags. The tags
  within the tag set are joined by a colon, e.g. '1:s//3:p', which is
  displayed as ''first-person singular and third-person plural'', e.g.
  for use with the form [[μέλλον]] of the verb [[μέλλω]] "to intend",
  which uses the tag set '1:s//3:p|impf|actv|indc|unaugmented' to express
  the syncretism between the first singular and third plural forms of the
  imperfect active indicative unaugmented conjugation. Two-level multipart
  tags should be used sparingly; if in doubt, list out the inflections
  separately.
* A MULTIPART TAG SHORTCUT is a shortcut that expands into a multipart
  tag, for example '123', which expands to the multipart tag '1//2//3'.
  Only the most common such combinations exist as shortcuts.
* A LIST TAG SHORTCUT is a special type of shortcut that expands to a list
  of tags instead of a single tag. For example, the shortcut '1s' expands to
  '1|s' (first-person singular). Only the most common such combinations
  exist as shortcuts.


]=]
* A ''tag'' is a single grammatical item, as specified in a single numbered parameter of {{tl|inflection of}}. Examples
  are `masculine`, `nominative`, or `first-person`. Tags may be abbreviated, e.g. `m` for `masculine`, `nom` for
  `nominative`, or `1` for `first-person`. Such abbreviations are called ''aliases'', and some tags have multiple
  equivalent aliases (e.g. `p` or `pl` for `plural`). The full, non-abbreviated form of a tag is called its
  ''canonical form''.
* The ''display form'' of a tag is the way it's displayed to the user. Usually the displayed text of the tag is the same
  as its canonical form, and it normally functions as a link to a glossary entry explaining the tag. Usually the link is
  to an entry in [[Appendix:Glossary]], but sometimes the tag is linked to an individual dictionary entry or to a
  Wikipedia entry. Occasionally, the display text differs from the canonical form of the tag. An example is the tag
  `comparative case`, which has the display text read as simply `comparative`. Normally, tags referring to cases don't
  have the word "case" in them, but in this case the tag `comparative` was already used as an alias for the tag
  `comparative degree`, so the tag was named `comparative case` to avoid clashing. A similar situation occurs with
  `adverbial case` vs. the grammar tag `adverbial` (as in `adverbial participle`).
* A ''tag set'' is an ordered list of tags, which together express a single inflection, for example, `1|s|pres|ind`,
  which can be expanded to canonical-form tags as `first-person|singular|present|indicative`.
* A ''conjoined tag set'' is a tag set that consists of multiple individual tag sets separated by a semicolon, e.g.
  `1|s|pres|ind|;|2|s|imp`, which specifies two tag sets, `1|s|pres|ind` as above and `2|s|imp` (in canonical form,
  `second-person|singular|imperative`). Multiple tag sets specified in a single call to {{tl|inflection of}} are
  specified in this fashion. Conjoined tag sets can also occur in list-tag shortcuts.
* A ''multipart tag'' is a tag that embeds multiple tags within it, such as `f//n` or `nom//acc//voc`. These are used in
  the case of [[syncretism]], when the same form applies to multiple inflections. Examples are the Spanish present
  subjunctive, where the first-person and third-person singular have the same form (e.g. {{m|es|siga}} from
  {{m|es|seguir|t=to follow}}), or Latin third-declension adjectives, where the dative and ablative plural of all
  genders have the same form (e.g. {{m|la|omnibus}} from {{m|la|omnis|t=all}}). These would be expressed respectively as
  `1//3|s|pres|sub` and `dat//abl|m//f//n|p`, where the use of the multipart tag compactly encodes the syncretism and
  avoids the need to individually list out all of the inflections. Multipart tags currently display as a list separated
  by a slash, e.g.  ''dative/ablative'' or ''masculine/feminine/neuter'' where each individual word is linked
  appropriately. As a special case, multipart tags involving persons display specially; for example, the multipart tag
  `1//2//3` displays as ''first-, second- and third-person'', with the word "person" occurring only once.
* A ''two-level multipart tag'' is a special type of multipart tag that joins two or more tag sets instead of joining
  individual tags. The tags within the tag set are joined by a colon, e.g. `1:s//3:p`, which is displayed as
  ''first-person singular and third-person plural'', e.g. for use with the form {{m|grc|μέλλον}} of the verb
  {{m|grc|μέλλω|t=to intend}}, which uses the tag set `1:s//3:p|impf|actv|indc|unaugmented` to express the syncretism
  between the first singular and third plural forms of the imperfect active indicative unaugmented conjugation.
  Two-level multipart tags should be used sparingly; if in doubt, list out the inflections separately. [FIXME: Make
  two-level multipart tags obsolete.]
* A ''shortcut'' is a tag that expands to any type of tag described above, or to any type of tag set described above.
  Aliases are a particular type of shortcut whose expansion is a single non-multipart tag.
* A ''multipart shortcut'' is a shortcut that expands into a multipart tag, for example `123`, which expands to the
  multipart tag `1//2//3`. Only the most common such combinations exist as shortcuts.
* A ''list shortcut'' is a special type of shortcut that expands to a list of tags instead of a single tag. For example,
  the shortcut `1s` expands to `1|s` (first-person singular). Only the most common such combinations exist as shortcuts.
* A ''conjoined shortcut'' is a special type of list shortcut that consists of a conjoined tag set (multiple logical tag
  sets). For example, the English language-specific shortcut `ed-form` expands to `spast|;|past|part`, expressing the
  common syncretism between simple past and past participle in English (and in this case, `spast` is itself a list
  shortcut that expands to `simple|past`).
]==]


-- version of rsubn() that discards all but the first return value
-- version of rsubn() that discards all but the first return value
Line 121: Line 129:




local function normalize_index(list, index)
if index < 0 then
return #list + index + 1
end
return index
end
-- FIXME, consider moving to [[Module:table]]
-- Return true if the list `tags1`, treated as a set, is a subset of the list `tags2`, also treated as a set.
local function is_subset(tags1, tags2)
tags1 = m_table.listToSet(tags1)
tags2 = m_table.listToSet(tags2)
for tag, _ in pairs(tags1) do
if not tags2[tag] then
return false
end
end
return true
end
-- FIXME, move to [[Module:table]]
local function slice(list, i, j)
--checkType("slice", 1, list, "table")
--checkType("slice", 2, i, "number", true)
--checkType("slice", 3, j, "number", true)
if i == nil then
i = 1
else
i = normalize_index(list, i)
end
j = normalize_index(list, j or -1)
local retval = {}
local k = 0
for index = i, j do
k = k + 1
retval[k] = list[index]
end
return retval
end
local function wrap_in_span(text, classes)
return ("<span class='%s'>%s</span>"):format(classes, text)
end
--[==[
Lowest-level implementation of form-of templates, including the general {{tl|form of}} as well as those that deal with
inflection tags, such as the general {{tl|inflection of}}, semi-specific variants such as {{tl|participle of}}, and
specific variants such as {{tl|past participle of}}. `data` contains all the information controlling the display, with
the following fields:
* `.text`: Text to insert before the lemmas. Wrapped in the value of `.text_classes`, or its default; see below.
* `.lemmas`: List of objects describing the lemma(s) of which the term in question is a non-lemma form. These are passed
  directly to {full_link()} in [[Module:links]]. Each object should have at minimum a `.lang` field containing the
  language of the lemma and a `.term` field containing the lemma itself. Each object is formatted using {full_link()}
  and then if there are more than one, they are joined using {serialCommaJoin()} in [[Module:table]]. Alternatively,
  `.lemmas` can be a string, which is displayed directly, or omitted, to show no lemma links and omit the connecting
  text.
* `.lemma_face`: "Face" to use when displaying the lemma objects. Usually should be set to {"term"}.
* `.enclitics`: List of enclitics to display after the lemmas, in parens.
* `.base_lemmas`: List of base lemmas to display after the lemmas, in the case where the lemmas in `.lemmas` are
  themselves forms of another lemma (the base lemma), e.g. a comparative, superlative or participle. Each object is of
  the form { { paramobj = PARAM_OBJ, lemmas = {LEMMA_OBJ, LEMMA_OBJ, ...} }} where PARAM_OBJ describes the properties
  of the base lemma parameter (i.e. the relationship between the intermediate and base lemmas) and LEMMA_OBJ is an
  object suitable to be passed to {full_link()} in [[Module:links]]. PARAM_OBJ is of the format
  { { param = "PARAM", tags = {"TAG", "TAG", ...} } where PARAM is the name of the parameter to {{tl|inflection of}}
  etc. that holds the base lemma(s) of the specified relationship and the tags describe the relationship, such as
  { {"comd"}} or { {"past", "part"}}.
* `.text_classes`: CSS classes used to wrap the tag text and lemma links. Default is {"form-of-definition use-with-mention"}
  for the tag text and lemma links, and additionally {"form-of-definition-link"} specifically for the lemma links.
  (FIXME: Should separate out the lemma links into their own field.)
* `.posttext`: Additional text to display after the lemma links.
]==]
function export.format_form_of(data)
function export.format_form_of(data)
if type(data) ~= "table" then
if type(data) ~= "table" then
error("First argument must now be a table of arguments")
error("Internal error: First argument must now be a table of arguments")
end
end
 
local text_classes = data.text_classes or "form-of-definition use-with-mention"
local text_classes = data.text_classes or "form-of-definition"
local lemma_classes = data.text_classes or "form-of-definition-link"
local terminfo_classes = data.text_classes or "form-of-definition-link"
local parts = {}
local parts = {}
table.insert(parts, "<span class='" .. text_classes .. "'>")
local function ins(text)
table.insert(parts, data.text)
table.insert(parts, text)
if data.text ~= "" and data.terminfo then
end
table.insert(parts, " ")
ins("<span class='" .. text_classes .. "'>")
ins(data.text)
if data.text ~= "" and data.lemmas then
ins(" ")
end
end
if data.terminfo then
if data.lemmas then
table.insert(parts, "<span class='" .. terminfo_classes .. "'>")
if type(data.lemmas) == "string" then
if type(data.terminfo) == "string" then
ins(wrap_in_span(data.lemmas, lemma_classes))
table.insert(parts, data.terminfo)
else
else
table.insert(parts, m_links.full_link(data.terminfo, data.terminfo_face, false))
local formatted_terms = {}
for _, lemma in ipairs(data.lemmas) do
table.insert(formatted_terms, wrap_in_span(
m_links.full_link(lemma, data.lemma_face), lemma_classes
))
end
ins(m_table.serialCommaJoin(formatted_terms))
end
end
if data.enclitics and #data.enclitics > 0 then
-- The outer parens need to be outside of the text_classes span so they show in upright instead of italic, or
-- they will clash with upright parens generated by link annotations such as transliterations and pos=.
ins("</span>")
local formatted_terms = {}
for _, enclitic in ipairs(data.enclitics) do
-- FIXME, should we have separate clitic face and/or classes?
table.insert(formatted_terms, wrap_in_span(
m_links.full_link(enclitic, data.lemma_face, nil, "show qualifiers"), lemma_classes
))
end
end
table.insert(parts, "</span>")
ins(" (")
ins(wrap_in_span("with enclitic" .. (#data.enclitics > 1 and "s" or "") .. " ", text_classes))
ins(m_table.serialCommaJoin(formatted_terms))
ins(")")
ins("<span class='" .. text_classes .. "'>")
end
end
if data.base_lemmas and #data.base_lemmas > 0 then
for _, base_lemma in ipairs(data.base_lemmas) do
ins(", the </span>")
ins(export.tagged_inflections {
lang = base_lemma.lemmas[1].lang,
tags = base_lemma.paramobj.tags,
lemmas = base_lemma.lemmas,
lemma_face = data.lemma_face,
no_format_categories = true,
nocat = true,
text_classes = data.text_classes,
})
ins("<span class='" .. text_classes .. "'>")
end
end
-- FIXME, should posttext go before enclitics? If so we need to have separate handling for the
-- final colon when there are multiple tag sets in tagged_inflections().
if data.posttext then
if data.posttext then
table.insert(parts, data.posttext)
ins(data.posttext)
end
end
table.insert(parts, "</span>")
ins("</span>")
return table.concat(parts)
return table.concat(parts)
end
end




local function is_link_or_html(tag)
--[==[
return tag:find("[[", nil, true) or tag:find("|", nil, true) or
Return true if `tag` contains an internal link or HTML.
tag:find("<", nil, true)
]==]
function export.is_link_or_html(tag)
return tag:find("[[", nil, true) or tag:find("|", nil, true) or tag:find("<", nil, true)
end
end




-- Look up a tag (either a shortcut of any sort of a canonical long-form tag)
--[==[
-- and return its expansion. The expansion will be a string unless the
Look up a tag (either a shortcut of any sort of a canonical long-form tag) and return its expansion. The expansion
-- shortcut is a list-tag shortcut such as "1s"; in that case, the expansion
will be a string unless the shortcut is a list-tag shortcut such as `1s`; in that case, the expansion will be a
-- will be a list. The caller must handle both cases. Only one level of
list. The caller must handle both cases. Only one level of expansion happens; hence, `acc` expands to {"accusative"},
-- expansion happens; hence, "acc" expands to "accusative", "1s" expands to
`1s` expands to { {"1", "s"}} (not to { {"first", "singular"}}) and `123` expands to {"1//2//3"}. The expansion will be
-- {"1", "s"} (not to {"first", "singular"}) and "123" expands to "1//2//3".
the same as the passed-in tag in the following circumstances:
-- The expansion will be the same as the passed-in tag in the following
 
-- circumstances:
# The tag is `;` (this is special-cased, and no lookup is done).
--
# The tag is a multipart tag such as `nom//acc` (this is special-cased, and no lookup is done).
-- 1. The tag is ";" (this is special-cased, and no lookup is done).
# The tag contains a raw link (this is special-cased, and no lookup is done).
-- 2. The tag is a multipart tag such as "nom//acc" (this is special-cased,
# The tag contains HTML (this is special-cased, and no lookup is done).
--    and no lookup is done).
# The tag is already a canonical long-form tag.
-- 3. The tag contains a raw link (this is special-cased, and no lookup is
# The tag is unrecognized.
--    done).
 
-- 4. The tag contains HTML (this is special-cased, and no lookup is done).
This function first looks up in the lang-specific data module [[Module:form of/lang-data/LANGCODE]], then in
-- 5. The tag is already a canonical long-form tag.
[[Module:form of/data]] (which includes more common non-lang-specific tags) and finally (only if the tag is not
-- 6. The tag is unrecognized.
recognized as a shortcut or canonical tag, and is not of types 1-4 above) in [[Module:form of/data2]].
--
 
-- This function first looks up in [[Module:form of/data]] (which includes
If the expansion is a string and is different from the tag, track it if `do_track` is true.
-- more common tags) and then (only if the tag is not recognized as a
]==]
-- shortcut or canonical tag, and is not of types 1-4 above) in
function export.lookup_shortcut(tag, lang, do_track)
-- [[Module:form of/data2]].
--
-- If the expansion is a string and is different from the tag, track it if
-- DO_TRACK is true.
function export.lookup_shortcut(tag)
-- If there is HTML or a link in the tag, return it directly; don't try
-- If there is HTML or a link in the tag, return it directly; don't try
-- to look it up, which will fail.
-- to look it up, which will fail.
if tag == ";" or tag:find("//", nil, true) or is_link_or_html(tag) then
if tag == ";" or tag:find("//", nil, true) or export.is_link_or_html(tag) then
return tag
return tag
end
end
local m_data = mw.loadData("Module:form of/data")
local expansion
-- If this is a canonical long-form tag, just return it, and don't
local langcode = lang and lang:getCode()
-- check for shortcuts (which will cause [[Module:form of/data2]] to be
if langcode and export.langs_with_lang_specific_tags[langcode] then
-- loaded).
local langdata = mw.loadData(export.form_of_lang_data_module_prefix .. langcode)
if m_data.tags[tag] then
-- If this is a canonical long-form tag, just return it, and don't check for shortcuts. This is an
return tag
-- optimization; see below.
if langdata.tags[tag] then
return tag
end
expansion = langdata.shortcuts[tag]
end
if not expansion and lang then
-- If the lang we're dealing with is an etym-only lang, try again with the corresponding full language.
local full_langcode = lang:getFullCode()
if full_langcode ~= langcode and export.langs_with_lang_specific_tags[full_langcode] then
local langdata = mw.loadData(export.form_of_lang_data_module_prefix .. full_langcode)
-- If this is a canonical long-form tag, just return it, and don't check for shortcuts. This is an
-- optimization; see below.
if langdata.tags[tag] then
return tag
end
expansion = langdata.shortcuts[tag]
end
end
if not expansion then
local m_data = mw.loadData(export.form_of_data_module)
-- If this is a canonical long-form tag, just return it, and don't check for shortcuts (which will cause
-- [[Module:form of/data2]] to be loaded, because there won't be a shortcut entry in [[Module:form of/data]] --
-- or, for that matter, in [[Module:form of/data2]]). This is an optimization; the code will still work without
-- it, but will use up more memory.
if m_data.tags[tag] then
return tag
end
expansion = m_data.shortcuts[tag]
end
end
local expansion = m_data.shortcuts[tag]
if not expansion then
if not expansion then
local m_data2 = mw.loadData("Module:form of/data2")
local m_data2 = mw.loadData(export.form_of_data2_module)
expansion = m_data2.shortcuts[tag]
expansion = m_data2.shortcuts[tag]
end
end
Line 203: Line 351:
return tag
return tag
end
end
return expansion
return expansion
end
end




-- Look up a normalized/canonicalized tag and return the data object
--[==[
-- associated with it. If the tag isn't found, return nil. This first looks up
Look up a normalized/canonicalized tag and return the data object associated with it. If the tag isn't found, return
-- in [[Module:form of/data]] (which includes more common tags) and then in
nil. This first looks up in the lang-specific data module [[Module:form of/lang-data/LANGCODE]], then in
-- [[Module:form of/data2]].
[[Module:form of/data]] (which includes more common non-lang-specific tags) and then finally in
function export.lookup_tag(tag)
[[Module:form of/data2]].
local m_data = mw.loadData("Module:form of/data")
]==]
function export.lookup_tag(tag, lang)
local langcode = lang and lang:getCode()
if langcode and export.langs_with_lang_specific_tags[langcode] then
local langdata = mw.loadData(export.form_of_lang_data_module_prefix .. langcode)
if langdata.tags[tag] then
return langdata.tags[tag]
end
end
local full_langcode = lang and lang:getFullCode()
if full_langcode and full_langcode ~= langcode and export.langs_with_lang_specific_tags[full_langcode] then
-- If the lang we're dealing with is an etym-only lang, try again with the corresponding full language.
local langdata = mw.loadData(export.form_of_lang_data_module_prefix .. full_langcode)
if langdata.tags[tag] then
return langdata.tags[tag]
end
end
local m_data = mw.loadData(export.form_of_data_module)
local tagobj = m_data.tags[tag]
local tagobj = m_data.tags[tag]
if tagobj then
if tagobj then
return tagobj
return tagobj
end
end
local m_data2 = mw.loadData("Module:form of/data2")
local m_data2 = mw.loadData(export.form_of_data2_module)
local tagobj2 = m_data2.tags[tag]
local tagobj2 = m_data2.tags[tag]
if tagobj2 then
if tagobj2 then
Line 226: Line 392:




-- Normalize a single tag, which may be a shortcut but should not be a
-- Normalize a single tag, which may be a shortcut but should not be a multipart tag, a multipart shortcut or a list
-- multipart tag, a multipart-tag shortcut or a list-tag shortcut.
-- shortcut.
local function normalize_single_tag(tag)
local function normalize_single_tag(tag, lang, do_track)
local expansion = export.lookup_shortcut(tag)
local expansion = export.lookup_shortcut(tag, lang, do_track)
if type(expansion) ~= "string" then
if type(expansion) ~= "string" then
error("Tag '" .. tag .. "' is a list-tag shortcut, which is not allowed here")
error("Tag '" .. tag .. "' is a list shortcut, which is not allowed here")
end
end
tag = expansion
tag = expansion
return tag
return tag
end
end




-- Normalize a component of a multipart tag. This should not have any // in it,
--[=[
-- but may join multiple individual tags with a colon, and may be a single
Normalize a component of a multipart tag. This should not have any // in it, but may join multiple individual tags with
-- list-tag shortcut, which is treates as if colon-separated. If
a colon, and may be a single list-tag shortcut, which is treated as if colon-separated. The return value may be a list
-- RECOMBINE_TAGS isn't given, the return value may be a list of tags;
of tags.
-- otherwise, it will always be a string, and multiple tags will be
]=]
-- represented as canonical-form tags joined by ":".
local function normalize_multipart_component(tag, lang, do_track)
local function normalize_multipart_component(tag, recombine_tags)
-- If there is HTML or a link in the tag, don't try to split on colon. A colon may legitimately occur in either one,
-- If there is HTML or a link in the tag, don't try to split on colon.
-- and we don't want these things parsed. Note that we don't do this check before splitting on //, which we don't
-- A colon may legitimately occur in either one, and we don't want
-- expect to occur in links or HTML; see comment in normalize_tag().
-- these things parsed. Note that we don't do this check before splitting
if export.is_link_or_html(tag) then
-- on //, which we don't expect to occur in links or HTML; see comment
-- in normalize_tag().
if is_link_or_html(tag) then
return tag
return tag
end
end
Line 258: Line 420:
-- We allow list-tag shortcuts inside of multipart tags, e.g.
-- We allow list-tag shortcuts inside of multipart tags, e.g.
-- '1s//3p'. Check for this now.
-- '1s//3p'. Check for this now.
tag = export.lookup_shortcut(tag)
tag = export.lookup_shortcut(tag, lang, do_track)
if type(tag) == "table" then
if type(tag) == "table" then
-- We found a list-tag shortcut; treat as if colon-separated.
-- We found a list-tag shortcut; treat as if colon-separated.
components = tag
components = tag
else
else
return normalize_single_tag(tag)
return normalize_single_tag(tag, lang, do_track)
end
end
end
end
local normtags = {}
local normtags = {}
for _, component in ipairs(components) do
for _, component in ipairs(components) do
table.insert(normtags, normalize_single_tag(component))
table.insert(normtags, normalize_single_tag(component, lang, do_track))
end
end


if recombine_tags then
return normtags
return table.concat(normtags, ":")
else
return normtags
end
end
end




-- Normalize a single tag. If RECOMBINE_TAGS isn't given, the return value
--[=[
-- may be a list (in the case of multipart tags), which will contain nested
Normalize a single tag. The return value may be a list (in the case of multipart tags), which will contain nested lists
-- lists in the case of two-level multipart tags; otherwise, it will always
in the case of two-level multipart tags.
-- be a string, and multipart tags will be represented as canonical-form tags
]=]
-- joined by "//" and/or ":".
local function normalize_tag(tag, lang, do_track)
local function normalize_tag(tag, recombine_multitags)
-- We don't check for links or HTML before splitting on //, which we don't expect to occur in links or HTML. Doing
-- We don't check for links or HTML before splitting on //, which we
-- it this way allows for a tag like '{{lb|grc|Epic}}//{{lb|grc|Ionic}}' to function correctly (the template calls
-- don't expect to occur in links or HTML. Doing it this way allows for
-- will be expanded before we process the tag, and will contain links and HTML). The only check we do is for a URL,
-- a tag like '{{lb|grc|Epic}}//{{lb|grc|Ionic}}' to function correctly
-- which shouldn't normally occur, but might if the user tries to put an external link into the tag. URL's with //
-- (the template calls will be expanded before we process the tag, and
-- normally have the sequence ://, which should never normally occur when // and : are used in their normal ways.
-- will contain links and HTML). The only check we do is for a URL,
-- which shouldn't normally occur, but might if the user tries to put
-- an external link into the tag. URL's with // normally have the
-- sequence ://, which should never normally occur when // and : are
-- used in their normal ways.
if tag:find("://", nil, true) then
if tag:find("://", nil, true) then
return tag
return tag
Line 299: Line 452:
local split_tags = rsplit(tag, "//", true)
local split_tags = rsplit(tag, "//", true)
if #split_tags == 1 then
if #split_tags == 1 then
local retval = normalize_multipart_component(tag, recombine_multitags)
local retval = normalize_multipart_component(tag, lang, do_track)
if type(retval) == "table" then
if type(retval) == "table" then
-- The user gave a tag like '1:s', i.e. with colon but without
-- The user gave a tag like '1:s', i.e. with colon but without //. Allow this, but we need to return a
-- //. Allow this, but we need to return a nested list. Note,
-- nested list.
-- this will never happen when RECOMBINE_TAGS is given.
return {retval}
return {retval}
end
end
Line 310: Line 462:
local normtags = {}
local normtags = {}
for _, single_tag in ipairs(split_tags) do
for _, single_tag in ipairs(split_tags) do
table.insert(normtags, normalize_multipart_component(single_tag,
table.insert(normtags, normalize_multipart_component(single_tag, lang, do_track))
recombine_multitags))
end
if recombine_multitags then
return table.concat(normtags, "//")
else
return normtags
end
end
return normtags
end
end




-- Normalize a tag set (a list of tags) into a list of canonical-form tags
--[==[
-- (which -- may be larger due to the possibility of list-tag shortcuts).
Normalize a tag set (a list of tags) into its canonical-form tags. The return value is a list of normalized tag sets
-- If RECOMBINE_TAGS isn't given, the return list may itself contains lists;
(a list because of there may be conjoined shortcuts among the input tags). A normalized tag set is a list of tag
-- in particular, multipart tags will be represented as lists. Specifically,
elements, where each element is either a string (the canonical form of a tag), a list of such strings (in the case of
-- the list will consist of the elements of the multipart tag, which will
multipart tags) or a list of lists of such strings (in the case of two-level multipart tags). For example, the multipart
-- either be canonical-form strings or (in the case of two-level multipart
tag `nom//acc//voc` will be represented in canonical form as { {"nominative", "accusative", "vocative"}}, and the
-- tags) nested lists of canonical-form strings. For example, the multipart
two-level multipart tag `1:s//3:p` will be represented as { {{"first-person", "singular"}, {"third-person", "plural"}}}.
-- tag ''nom//acc//voc'' will expand to
 
--  {"nominative", "accusative", "vocative"}
Example 1:
-- and the two-level multipart tag ''1:s//3:p'' will expand to
 
--  {{"first-person", "singular"}, {"third-person", "plural"}}.
{normalize_tag_set({"nom//acc//voc", "n", "p"})} = { {{{"nominative", "accusative", "vocative"}, "masculine", "plural"}}}
-- If RECOMBINE_TAGS is given, multipart tags will be represented in string
 
-- form, i.e. as canonical-form tags joined by "//" and/or ":".
Example 2:
function export.normalize_tags(tags, recombine_multitags)
 
local ntags = {}
{normalize_tag_set({"ed-form"}, ENGLISH)} = { {{"simple", "past"}, {"past", "participle"}}}
for _, tag in ipairs(tags) do
 
-- Expand the tag, which may generate a new tag (either a
Example 3:
-- fully canonicalized tag, a multipart tag, or a list of tags).
 
tag = export.lookup_shortcut(tag)
{normalize_tag_set({"archaic", "ed-form"}, ENGLISH)} = { {{"archaic", "simple", "past"}, {"archaic", "past", "participle"}}}
]==]
function export.normalize_tag_set(tag_set, lang, do_track)
local output_tag_set = {}
local saw_semicolon = false
 
for _, tag in ipairs(tag_set) do
-- Expand the tag, which may generate a new tag (either a fully canonicalized tag, a multipart tag, or a list
-- of tags).
tag = export.lookup_shortcut(tag, lang, do_track)
if type(tag) == "table" then
if type(tag) == "table" then
saw_semicolon = m_table.contains(tag, ";")
if saw_semicolon then
-- If we saw a conjoined shortcut, we need to use a more general algorithm that can expand a single
-- tag set into multiple.
break
end
for _, t in ipairs(tag) do
for _, t in ipairs(tag) do
table.insert(ntags, normalize_tag(t, recombine_multitags))
table.insert(output_tag_set, normalize_tag(t, lang, do_track))
end
else
table.insert(output_tag_set, normalize_tag(tag, lang, do_track))
end
end
 
if not saw_semicolon then
return {output_tag_set}
end
 
-- Use a more general algorithm that handles conjoined shortcuts.
local output_tag_set = {}
for i, tag in ipairs(tag_set) do
-- Expand the tag, which may generate a new tag (either a fully canonicalized tag, a multipart tag, or a list
-- of tags).
tag = export.lookup_shortcut(tag, lang, do_track)
if type(tag) == "table" then
local output_tag_sets = {}
local shortcut_tag_sets = export.split_tag_set(tag)
local normalized_shortcut_tag_sets = {}
for _, shortcut_tag_set in ipairs(shortcut_tag_sets) do
m_table.extendList(normalized_shortcut_tag_sets,
export.normalize_tag_set(shortcut_tag_set, lang, do_track))
end
end
local after_tags = slice(tag_set, i + 1)
local normalized_after_tags_sets = export.normalize_tag_set(after_tags, lang, do_track)
for _, normalized_shortcut_tag_set in ipairs(normalized_shortcut_tag_sets) do
for _, normalized_after_tags_set in ipairs(normalized_after_tags_sets) do
table.insert(output_tag_sets, m_table.append(output_tag_set, normalized_shortcut_tag_set,
normalized_after_tags_set))
end
end
return output_tag_sets
else
else
table.insert(ntags, normalize_tag(tag, recombine_multitags))
table.insert(output_tag_set, normalize_tag(tag, lang, do_track))
end
end
end
end
return ntags
 
error("Internal error: Should not get here")
end
end




-- Split a tag set containing two-level multipart tags into one or more tag sets not containing such tags.
function export.combine_multipart_tags(tag_set)
-- Single-level multipart tags are left alone. (If we need to, a slight modification of the following code
for i, tag in ipairs(tag_set) do
-- will also split single-level multipart tags.) This assumes that multipart tags are represented as lists
if type(tag) == "table" then
-- and two-level multipart tags are represented as lists of lists, as is output by normalize_tags().
for j, subtag in ipairs(tag) do
-- NOTE: We have to be careful to properly handle imbalanced two-level multipart tags such as
if type(subtag) == "table" then
-- <code>def:s//p</code> (or the reverse, <code>s//def:p</code>).
tag[j] = table.concat(subtag, ":")
end
end
tag_set[i] = table.concat(tag, "//")
end
end
 
return tag_set
end
 
 
function export.normalize_tags(tags, lang, recombine_multitags, do_track)
local tag_sets = export.normalize_tag_set(tags, lang, do_track)
if recombine_multitags then
for i, tag_set in ipairs(tag_sets) do
tag_sets[i] = export.combine_multipart_tags(tag_set)
end
return export.combine_tag_sets(tag_sets)
end
return tag_sets
end
 
 
--[==[
Split a tag set containing two-level multipart tags into one or more tag sets not containing such tags.
Single-level multipart tags are left alone. (If we need to, a slight modification of the following code
will also split single-level multipart tags.) This assumes that multipart tags are represented as lists
and two-level multipart tags are represented as lists of lists, as is output by {normalize_tag_set()}.
NOTE: We have to be careful to properly handle imbalanced two-level multipart tags such as
`def:s//p` (or the reverse, `s//def:p`).
]==]
function export.split_two_level_multipart_tag_set(tag_set)
function export.split_two_level_multipart_tag_set(tag_set)
-- This would be a whole lot easier in Python, with built-in support for
-- slicing and array concatenation.
for i, tag in ipairs(tag_set) do
for i, tag in ipairs(tag_set) do
if type(tag) == "table" then
if type(tag) == "table" then
Line 374: Line 599:
-- We found a two-level multipart tag.
-- We found a two-level multipart tag.
-- (1) Extract the preceding tags.
-- (1) Extract the preceding tags.
local pre_tags = {}
local pre_tags = slice(tag_set, 1, i - 1)
for j=1,i-1 do
table.insert(pre_tags, tag_set[j])
end
-- (2) Extract the following tags.
-- (2) Extract the following tags.
local post_tags = {}
local post_tags = slice(tag_set, i + 1)
for j=i+1,#tag_set do
table.insert(post_tags, tag_set[j])
end
-- (3) Loop over each tag set alternant in the two-level multipart tag.
-- (3) Loop over each tag set alternant in the two-level multipart tag.
-- For each alternant, form the tag set consisting of pre_tags + alternant + post_tags,
-- For each alternant, form the tag set consisting of pre_tags + alternant + post_tags,
Line 389: Line 608:
for _, first_level_tag_set in ipairs(tag) do
for _, first_level_tag_set in ipairs(tag) do
local expanded_tag_set = {}
local expanded_tag_set = {}
for _, pre_tag in ipairs(pre_tags) do
m_table.extendList(expanded_tag_set, pre_tags)
table.insert(expanded_tag_set, pre_tag)
end
-- The second level may have a string or a list.
-- The second level may have a string or a list.
if type(first_level_tag_set) == "table" then
if type(first_level_tag_set) == "table" then
for _, second_level_tag in ipairs(first_level_tag_set) do
m_table.extendList(expanded_tag_set, first_level_tag_set)
table.insert(expanded_tag_set, second_level_tag)
end
else
else
table.insert(expanded_tag_set, first_level_tag_set)
table.insert(expanded_tag_set, first_level_tag_set)
end
end
for _, post_tag in ipairs(post_tags) do
m_table.extendList(expanded_tag_set, post_tags)
table.insert(expanded_tag_set, post_tag)
m_table.extendList(resulting_tag_sets, export.split_two_level_multipart_tag_set(expanded_tag_set))
end
for _, split_tag_set in ipairs(export.split_two_level_multipart_tag_set(expanded_tag_set)) do
table.insert(resulting_tag_sets, split_tag_set)
end
end
end
return resulting_tag_sets
return resulting_tag_sets
Line 416: Line 627:




-- Given a list of tags, split into tag sets (separated by semicolons in the initial list of tags).
--[==[
function export.split_tags_into_tag_sets(tags)
Split a tag set that may consist of multiple semicolon-separated tag sets into the component tag sets.
local tag_set_group = {}
]==]
function export.split_tag_set(tag_set)
local split_tag_sets = {}
local cur_tag_set = {}
local cur_tag_set = {}
for _, tag in ipairs(tags) do
for _, tag in ipairs(tag_set) do
if tag == ";" then
if tag == ";" then
if #cur_tag_set > 0 then
if #cur_tag_set > 0 then
table.insert(tag_set_group, cur_tag_set)
table.insert(split_tag_sets, cur_tag_set)
end
end
cur_tag_set = {}
cur_tag_set = {}
Line 431: Line 644:
end
end
if #cur_tag_set > 0 then
if #cur_tag_set > 0 then
table.insert(tag_set_group, cur_tag_set)
table.insert(split_tag_sets, cur_tag_set)
end
end
return tag_set_group
return split_tag_sets
end
end
export.split_tags_into_tag_sets = export.split_tag_set




-- Given a list of tags, split into tag sets (separated by semicolons in the initial list of tags).
--[==[
-- Then, potentially split each tag set into multiple tag sets if there are any two-level multipart
Combine multiple tag sets in a tag set group into a simple tag set, with logical tag sets separated by semicolons.
-- tags in those tag sets.
This is the opposite of {split_tag_set()}.
function export.split_tags_into_tag_sets_and_expand_two_level_multipart_tags(tags)
]==]
-- First, split into tag sets.
function export.combine_tag_sets(tag_sets)
local tag_sets = export.split_tags_into_tag_sets(tags)
if #tag_sets == 1 then
-- Now split any two-level multipart tags.
return tag_sets[1]
local resulting_tag_sets = {}
end
local combined_tag_set = {}
for _, tag_set in ipairs(tag_sets) do
for _, tag_set in ipairs(tag_sets) do
for _, resulting_tag_set in ipairs(export.split_two_level_multipart_tag_set(tag_set)) do
if #combined_tag_set > 0 then
table.insert(resulting_tag_sets, resulting_tag_set)
table.insert(combined_tag_set, ";")
end
end
m_table.extendList(combined_tag_set, tag_set)
end
return tags
end
local tag_set_param_mods = {
lb = {
item_dest = "labels",
convert = function(arg, parse_err)
return rsplit(arg, "//", true)
end,
}
}
--[==[
Parse tag set properties from a tag set (list of tags). Currently no per-tag properties are recognized, and the only
per-tag-set property recognized is `<lb:...>` for specifing label(s) for the tag set. Per-tag-set properties must be
attached to the last tag.
]==]
function export.parse_tag_set_properties(tag_set)
local function generate_tag_set_obj(last_tag)
tag_set[#tag_set] = last_tag
return {tags = tag_set}
end
local last_tag = tag_set[#tag_set]
-- Check for inline modifier, e.g. מרים<tr:Miryem>. But exclude HTML entry with <span ...>, <i ...>, <br/> or
-- similar in it, caused by wrapping an argument in {{l|...}}, {{af|...}} or similar. Basically, all tags of
-- the sort we parse here should consist of a less-than sign, plus letters, plus a colon, e.g. <lb:...>, so if
-- we see a tag on the outer level that isn't in this format, we don't try to parse it. The restriction to the
-- outer level is to allow generated HTML inside of e.g. qualifier tags, such as foo<q:similar to {{m|fr|bar}}>.
if last_tag:find("<") and not last_tag:find("^[^<]*<[a-z]*[^a-z:]") then
return require(parse_utilities_module).parse_inline_modifiers(last_tag, {
param_mods = tag_set_param_mods,
generate_obj = generate_tag_set_obj,
})
else
return generate_tag_set_obj(last_tag)
end
end
return resulting_tag_sets
end
end




function export.normalize_pos(pos)
function export.normalize_pos(pos)
return m_pos[pos] or pos
if not pos then
return nil
end
return mw.loadData(export.form_of_pos_module)[pos] or pos
end
end


Line 462: Line 719:
-- passed in must be a string (i.e. it cannot be a list describing a
-- passed in must be a string (i.e. it cannot be a list describing a
-- multipart tag). To handle multipart tags, use get_tag_display_form().
-- multipart tag). To handle multipart tags, use get_tag_display_form().
local function get_single_tag_display_form(normtag)
local function get_single_tag_display_form(normtag, lang)
local data = export.lookup_tag(normtag)
local data = export.lookup_tag(normtag, lang)
local display = normtag


-- If the tag has a special display form, use it
-- If the tag has a special display form, use it
if data and data.display then
if data and data.display then
normtag = data.display
display = data.display
end
end


-- If there is a nonempty glossary index, then show a link to it
-- If there is a nonempty glossary index, then show a link to it
if data and data.glossary then
local glossary = data and data[export.GLOSSARY]
if data.glossary_type == "wikt" then
if glossary ~= nil then
normtag = "[[" .. data.glossary .. "|" .. normtag .. "]]"
if glossary == export.WIKT then
elseif data.glossary_type == "wp" then
display = "[[" .. normtag .. "|" .. display .. "]]"
normtag = "[[w:" .. data.glossary .. "|" .. normtag .. "]]"
elseif glossary == export.WP then
display = "[[w:" .. normtag .. "|" .. display .. "]]"
elseif glossary == export.APPENDIX then
display = "[[Appendix:Glossary#" .. mw.uri.anchorEncode(normtag) .. "|" .. display .. "]]"
elseif type(glossary) ~= "string" then
error(("Internal error: Wrong type %s for glossary value %s for tag %s"):format(
type(glossary), mw.dumpObject(glossary), normtag))
else
else
normtag = "[[wikt:Appendix:Glossary#" .. mw.uri.anchorEncode(data.glossary) .. "|" .. normtag .. "]]"
local link = rmatch(glossary, "^wikt:(.*)")
if link then
display = "[[" .. link .. "|" .. display .. "]]"
end
if not link then
link = rmatch(glossary, "^w:(.*)")
if link then
display = "[[w:" .. link .. "|" .. display .. "]]"
end
end
if not link then
display = "[[Appendix:Glossary#" .. mw.uri.anchorEncode(glossary) .. "|" .. display .. "]]"
end
end
end
end
end
return normtag
return display
end
end




-- Turn a canonicalized tag spec (which describes a single, possibly
--[==[
-- multipart tag) into the displayed form. The tag spec may be a string
Turn a canonicalized tag spec (which describes a single, possibly multipart tag) into the displayed form. The tag spec
-- (a canonical-form tag), or a list of canonical-form tags (in the
may be a string (a canonical-form tag); a list of canonical-form tags (in the case of a simple multipart tag); or a
-- case of a simple multipart tag), or a list of mixed canonical-form
list of mixed canonical-form tags and lists of such tags (in the case of a two-level multipart tag). `joiner` indicates
-- tags and lists of such tags (in the case of a two-level multipart tag).
how to join the parts of a multipart tag, and can be either {"and"} ("foo and bar", or "foo, bar and baz" for 3 or
-- JOINER indicates how to join the parts of a multipart tag, and can
more), {"slash"} ("foo/bar"), {"en-dash"} ("foo–bar") or {nil}, which uses the global default found in
-- be either "and" ("foo and bar", or "foo, bar and baz" for 3 or more),
{multipart_join_strategy()} in [[Module:form of/functions]]. (NOTE: The global default is {"slash"} and this seems
-- "slash" ("foo/bar"), "en-dash" ("foo–bar") or nil, which uses the
unlikely to change.)
-- global default found in multipart_join_strategy() in
]==]
-- [[Module:form of/functions]].
function export.get_tag_display_form(tagspec, lang, joiner)
function export.get_tag_display_form(tagspec, joiner)
if type(tagspec) == "string" then
if type(tagspec) == "string" then
return get_single_tag_display_form(tagspec)
return get_single_tag_display_form(tagspec, lang)
end
end
-- We have a multipart tag. See if there's a display handler to
-- We have a multipart tag. See if there's a display handler to display them specially.
-- display them specially.
for _, handler in ipairs(require(export.form_of_functions_module).display_handlers) do
for _, handler in ipairs(m_functions.display_handlers) do
local displayval = handler(tagspec, joiner)
local displayval = handler(tagspec, joiner)
if displayval then
if displayval then
Line 510: Line 784:
for _, first_level_tag in ipairs(tagspec) do
for _, first_level_tag in ipairs(tagspec) do
if type(first_level_tag) == "string" then
if type(first_level_tag) == "string" then
table.insert(displayed_tags, get_single_tag_display_form(first_level_tag))
table.insert(displayed_tags, get_single_tag_display_form(first_level_tag, lang))
else
else
-- A first-level element of a two-level multipart tag.
-- A first-level element of a two-level multipart tag. Currently we just separate the individual components
-- Currently we just separate the individual components
-- with spaces, but other ways are possible, e.g. using an underscore, colon, parens or braces.
-- with spaces, but other ways are possible, e.g. using
-- an underscore, colon, parens or braces.
local components = {}
local components = {}
for _, component in ipairs(first_level_tag) do
for _, component in ipairs(first_level_tag) do
table.insert(components, get_single_tag_display_form(component))
table.insert(components, get_single_tag_display_form(component, lang))
end
end
table.insert(displayed_tags, table.concat(components, " "))
table.insert(displayed_tags, table.concat(components, " "))
end
end
end
end
return m_functions.join_multiparts(displayed_tags, joiner)
return require(export.form_of_functions_module).join_multiparts(displayed_tags, joiner)
end
end




-- Return true if the list `tags1`, treated as a set, is a subset of the list `tags2`, also
--[==[
-- treated as a set.
Given a normalized tag set (i.e. as output by {normalize_tag_set()}; all tags are in canonical form, multipart tags are
local function is_subset(tags1, tags2)
represented as lists, and two-level multipart tags as lists of lists), convert to displayed form (a string). See
tags1 = m_table.listToSet(tags1)
{get_tag_display_form()} for the meaning of `joiner`.
tags2 = m_table.listToSet(tags2)
]==]
for tag, _ in pairs(tags1) do
function export.get_tag_set_display_form(normalized_tag_set, lang, joiner)
if not tags2[tag] then
local parts = {}
return false
 
for _, tagspec in ipairs(normalized_tag_set) do
local to_insert = export.get_tag_display_form(tagspec, lang, joiner)
-- Maybe insert a space before inserting the display form of the tag. We insert a space if
-- (a) we're not the first tag; and
-- (b) the tag we're about to insert doesn't have the "no_space_on_left" property; and
-- (c) the preceding tag doesn't have the "no_space_on_right" property.
-- NOTE: We depend here on the fact that
-- (1) all tags with either of the above properties set have the same display form as canonical form, and
-- (2) all tags with either of the above properties set are single-character tags.
-- The second property is an optimization to avoid looking up display forms resulting from multipart tags,
-- which won't be found and which will trigger loading of [[Module:form of/data2]]. If multichar punctuation is
-- added in the future, it's ok to change the == 1 below to <= 2 or <= 3.
--
-- If the first property above fails to hold in the future, we need to track the canonical form of each tag
-- (including the previous one) as well as the display form. This would also avoid the need for the == 1 check.
if #parts > 0 then
local most_recent_tagobj = ulen(parts[#parts]) == 1 and export.lookup_tag(parts[#parts], lang)
local to_insert_tagobj = ulen(to_insert) == 1 and export.lookup_tag(to_insert, lang)
if (
(not most_recent_tagobj or not most_recent_tagobj.no_space_on_right) and
(not to_insert_tagobj or not to_insert_tagobj.no_space_on_left)
) then
table.insert(parts, " ")
end
end
end
table.insert(parts, to_insert)
end
end
return true
 
return table.concat(parts)
end
end




-- Compute and return the appropriate categories for the tags in `tags` (user-specified tags,
--[==[
-- which may consist of multiple tag sets separated by semicolons) and the language in `lang`.
Given a normalized tag set (i.e. as output by {normalize_tag_set()}; all tags are in canonical form, multipart tags are
-- This checks both language-specific and language-agnostic category specs in [[Module:form of/cats]].
represented as lists, and two-level multipart tags as lists of lists), fetch the associated categories and labels.
-- `POS` is the user-specified part of speech, if any, and `terminfo` is currently unused.
Return two values, a list of categories and a list of labels. `lang` is the language of term represented by the tag set,
function export.fetch_lang_categories(lang, tags, terminfo, POS)
and `POS` is the user-provided part of speech (which may be {nil}).
local m_cats = mw.loadData("Module:form of/cats")
]==]
 
function export.fetch_categories_and_labels(normalized_tag_set, lang, POS, pagename, lemmas)
local m_cats = mw.loadData(export.form_of_cats_module)
local categories = {}
local categories = {}
local labels = {}


local normalized_tags = export.normalize_tags(tags)
local split_tag_sets = export.split_tags_into_tag_sets_and_expand_two_level_multipart_tags(normalized_tags)
POS = export.normalize_pos(POS)
POS = export.normalize_pos(POS)
 
-- First split any two-level multipart tags into multiple sets, to make our life easier.
-- Loop over each tag set and compute categories for each one.
for _, tag_set in ipairs(export.split_two_level_multipart_tag_set(normalized_tag_set)) do
for _, tag_set in ipairs(split_tag_sets) do
-- Call a named function, either from the lang-specific data in
local function make_function_table()
-- [[Module:form of/lang-specific/LANGCODE/functions]] or in [[Module:form of/functions]].
return {
local function call_named_function(name, funtype)
lang=lang,
local data = {
tags=normalized_tags,
pagename = pagename or mw.title.getCurrentTitle().subpageText,
term=term,
lemmas = lemmas,
p=POS
tag_set = normalized_tag_set,
lang = lang,
POS = POS
}
}
local modules_tried = {}
local function try_lang_specific_module(langcode)
if export.langs_with_lang_specific_tags[langcode] then
local lang_specific_module = export.form_of_lang_data_module_prefix .. langcode .. "/functions"
local langdata = require(utilities_module).safe_require(lang_specific_module)
if langdata then
table.insert(modules_tried, lang_specific_module)
if langdata.cat_functions then
local fn = langdata.cat_functions[name]
if fn then
return fn(data), true
end
end
end
end
return nil, false
end
-- First try lang-specific.
local langcode = lang and lang:getCode()
if langcode then
local retval, found_it = try_lang_specific_module(langcode)
if found_it then
return retval
end
end
-- If the lang we're dealing with is an etym-only lang, try again with the corresponding full language.
local full_langcode = lang and lang:getFullCode()
if full_langcode and full_langcode ~= langcode then
local retval, found_it = try_lang_specific_module(full_langcode)
if found_it then
return retval
end
end
-- Try lang-independent.
table.insert(modules_tried, export.form_of_functions_module)
local fn = require(export.form_of_functions_module).cat_functions[name]
if fn then
return fn(data)
end
for i, modname in ipairs(modules_tried) do
modules_tried[i] = "[[" .. modname .. "]]"
end
error(("No %s function named '%s' in %s"):format(funtype, name, lang_specific_part,
m_table.serialCommaJoin(modules_tried, {conj = "or", dontTag = true})))
end
end


Line 599: Line 943:
-- complex when multipart tags are present.
-- complex when multipart tags are present.
local function tag_set_matches_spec_tag(spec_tag)
local function tag_set_matches_spec_tag(spec_tag)
spec_tag = normalize_tag(spec_tag)
spec_tag = normalize_tag(spec_tag, lang)
for _, tag_set_tag in ipairs(tag_set) do
for _, tag_set_tag in ipairs(tag_set) do
if tag_set_tag_matches_spec_tag(tag_set_tag, spec_tag) then
if tag_set_tag_matches_spec_tag(tag_set_tag, spec_tag) then
Line 635: Line 979:
return false, 3
return false, 3
elseif predicate == "tags=" then
elseif predicate == "tags=" then
local normalized_spec_tags = export.normalize_tags(spec[2])
local normalized_spec_tag_sets = export.normalize_tag_set(spec[2], lang)
-- Allow tags to be in different orders, and multipart tags to
if #normalized_spec_tag_sets > 1 then
-- be in different orders. To handle this, we first check that
error("Internal error: No support for conjoined shortcuts in category/label specs in "
-- both tag set tags and spec tags have the same length. If so,
.. "[[Module:form of/cats]] when processing spec tag set " .. table.concat(spec[2], "|"))
-- we sort the multipart tags in the tag set tags and spec tags,
end
-- and then check that all tags in the spec tags are in the
local normalized_spec_tag_set = normalized_spec_tag_sets[1]
-- tag set tags.
-- Check for and disallow two-level multipart tags in the specs. FIXME: Remove this when we remove
if #tag_set ~= #normalized_spec_tags then
-- support for two-level multipart tags.
for _, tag in ipairs(normalized_spec_tag_set) do
if type(tag) == "table" then
for _, subtag in ipairs(tag) do
if type(subtag) == "table" then
error("Internal error: No support for two-level multipart tags in category/label specs"
.. "[[Module:form of/cats]] when processing spec tag set "
.. table.concat(spec[2], "|"))
end
end
end
end
-- Allow tags to be in different orders, and multipart tags to be in different orders. To handle this,
-- we first check that both tag set tags and spec tags have the same length. If so, we sort the
-- multipart tags in the tag set tags and spec tags, and then check that all tags in the spec tags are
-- in the tag set tags.
if #tag_set ~= #normalized_spec_tag_set then
return false, 3
return false, 3
end
end
Line 650: Line 1,010:
table.sort(tag_set_tags[i])
table.sort(tag_set_tags[i])
end
end
if type(normalized_spec_tags[i]) == "table" then
if type(normalized_spec_tag_set[i]) == "table" then
table.sort(normalized_spec_tags[i])
table.sort(normalized_spec_tag_set[i])
end
end
end
end
for i=1,#tag_set_tags do
for i=1,#tag_set_tags do
if not m_table.contains(tag_set_tags, normalized_spec_tags[i], "deepCompare") then
if not m_table.contains(tag_set_tags, normalized_spec_tag_set[i]) then
return false, 3
return false, 3
end
end
Line 686: Line 1,046:
end
end
return condval, 4
return condval, 4
elseif predication == "call" then
elseif predicate == "call" then
local fn = m_functions.cat_functions[spec[2]]
return fn(call_named_function(spec[2], "condition")), 3
if not fn then
error("No condition function named '" .. spec[2] .. "'")
end
return fn(make_function_table()), 3
else
else
error("Unrecognized predicate: " .. predicate)
error("Unrecognized predicate: " .. predicate)
Line 706: Line 1,062:
return false
return false
elseif type(spec) == "string" then
elseif type(spec) == "string" then
-- Substitute POS request with user-specified part of speech
-- A category. Substitute POS request with user-specified part of speech or default.
-- or default
spec = rsub(spec, "<<p=(.-)>>", function(default)
spec = rsub(spec, "<<p=(.-)>>", function(default)
return POS or export.normalize_pos(default)
return POS or export.normalize_pos(default)
end)
end)
table.insert(categories, lang:getCanonicalName() .. " " .. spec)
table.insert(categories, lang:getFullName() .. " " .. spec)
return true
elseif type(spec) == "table" and spec.labels then
-- A label spec.
for _, label in ipairs(spec.labels) do
m_table.insertIfNot(labels, label)
end
return true
return true
elseif type(spec) ~= "table" then
elseif type(spec) ~= "table" then
Line 734: Line 1,095:
return false
return false
elseif predicate == "call" then
elseif predicate == "call" then
local fn = m_functions.cat_functions[spec[2]]
return process_spec(call_named_function(spec[2], "spec"))
if not fn then
error("No spec function named '" .. spec[2] .. "'")
end
return process_spec(fn(make_function_table()))
else
else
local condval, ifspec = check_condition(spec)
local condval, ifspec = check_condition(spec)
Line 752: Line 1,109:
end
end


local langspecs = m_cats[lang:getCode()]
local langcode = lang:getCode()
local langspecs = m_cats[langcode]
if langspecs then
if langspecs then
for _, spec in ipairs(langspecs) do
for _, spec in ipairs(langspecs) do
Line 758: Line 1,116:
end
end
end
end
if lang:getCode() ~= "und" then
local full_code = lang:getFullCode()
if full_code ~= langcode then
local langspecs = m_cats[full_code]
if langspecs then
for _, spec in ipairs(langspecs) do
process_spec(spec)
end
end
end
if full_code ~= "und" then
local langspecs = m_cats["und"]
local langspecs = m_cats["und"]
if langspecs then
if langspecs then
Line 768: Line 1,135:
end
end


return categories
return categories, labels
end
end




function export.tagged_inflections(data, terminfo, notext, capfirst, posttext, joiner)
--[==[
if not data.tags then
Implementation of templates that display inflection tags, such as the general {{tl|inflection of}}, semi-specific
error("First argument must now be a table of arguments")
variants such as {{tl|participle of}}, and specific variants such as {{tl|past participle of}}. `data` contains all the
information controlling the display, with the following fields:
 
* `.lang`: ('''''required''''') Language to use when looking up language-specific inflection tags, categories and
  labels, and for displaying categories and labels.
* `.tags`: ('''''required''' unless `.tag_sets` is given'') List of non-canonicalized inflection tags. Multiple tag sets
  can be indicated by a {";"} as one of the tags, and tag-set properties may be attached to the last tag of a tag set.
  The tags themselves may come directly from the user (as in {{tl|inflection of}}); come partly from the user (as in
  {{tl|participle of}}, which adds the tag `part` to user-specified inflection tags); or be entirely specified by the
  template (as in {{tl|past participle of}}).
* `.tag_sets`: ('''''required''' unless `.tags` is given'') List of non-canonicalized tag sets and associated
  per-tag-set properties. Each element of the list is an object of the form
  { {tags = {"TAG", "TAG", ...}, labels = {"LABEL", "LABEL", ...}}. If `.tag_sets` is specified, `.tags` should not be
  given and vice-versa. Specifying `.tag_sets` in place of tags allowed per-tag set labels to be specified; otherwise,
  there is no advantage. [[Module:pt-gl-inflections]] uses this functionality to supply labels like {"Brazil"} and
  {"Portugal"} associated with specific tag sets.
* `.lemmas`: ('''''recommended''''') List of objects describing the lemma(s) of which the term in question is a
  non-lemma form. These are passed directly to {full_link()} in [[Module:links]]. Each object should have at minimum a
  `.lang` field containing the language of the lemma and a `.term` field containing the lemma itself. Each object is
  formatted using {full_link()} and then if there are more than one, they are joined using {serialCommaJoin()} in
  [[Module:table]]. Alternatively, `.lemmas` can be a string, which is displayed directly. If omitted entirely, no lemma
  links are shown and the connecting "of" is also omitted.
* `.lemma_face`: ('''''recommended''''') "Face" to use when displaying the lemma objects. Usually should be set to
  {"term"}.
* `.POS`: ('''''recommended''''') Categorizing part-of-speech tag. Comes from the {{para|p}} or {{para|POS}} argument of
  {{tl|inflection of}}.
* `.pagename`: Page name of "current" page or nil to use the actual page title; for testing purposes.
* `.enclitics`: List of enclitics to display after the lemmas, in parens.
* `.no_format_categories`: If true, don't format the categories derived from the inflection tags; just return them.
* `.sort`: Sort key for formatted categories. Ignored when `.no_format_categories` = {true}.
* `.nocat`: Suppress computation of categories (even if `.no_format_categories` is not given).
* `.notext`: Disable display of all tag text and `inflection of` text. (FIXME: Maybe not implemented correctly.)
* `.capfirst`: Capitalize the first word displayed.
* `.pretext`: Additional text to display before the inflection tags, but after any top-level labels.
* `.posttext`: Additional text to display after the lemma links.
* `.text_classes`: CSS classes used to wrap the tag text and lemma links. Default is
  {"form-of-definition use-with-mention"} for the tag text, {"form-of-definition-link"} for the lemma links. (FIXME:
  Should separate out the lemma links into their own field.)
`.joiner`: Override the joiner (normally a slash) used to join multipart tags. You should normally not specify this.
 
A typical call might look like this (for {{m+|es|amo}}): {
local lang = require("Module:languages").getByCode("es")
 
local lemma_obj = {
lang = lang,
term = "amar",
}
 
return m_form_of.tagged_inflections({
lang = lang, tags = {"1", "s", "pres", "ind"}, lemmas = {lemma_obj}, lemma_face = "term", POS = "verb"
})
}
 
Normally, one value is returned, the formatted text, which has appended to it the formatted categories derived from the
tag-set-related categories generated by the specs in [Module:form of/cats]]. To suppress this, set
`data.no_format_categories` = {true}, in which case two values are returned, the formatted text without any formatted
categories appended and a list of the categories to be formatted.
 
NOTE: There are two sets of categories that may be generated: (1) categories derived directly from the tag sets, as
specified in [[Module:form of/cats]]; (2) categories derived from tag-set labels, either (a) set explicitly by the
caller in `data.tag_sets`, (b) specified by the user using `<lb:...>` attached to the last tag in a tag set, or
(c) specified in [[Module:form of/cats]]. The second type (label-related categories) are currently not returned in
the second return value of {tagged_inflections()}, and are currently inserted into the output text even if
`data.no_format_categories` is set to {true}; but they can be suppressed by setting `data.nocat` = {true} (which also
suppresses the first type of categories, those derived directly from tag sets, even if `data.no_format_categories` is
set to {true}).
]==]
function export.tagged_inflections(data)
if not data.tags and not data.tag_sets then
error("First argument must be a table of arguments, and `.tags` or `.tag_sets` must be specified")
end
if data.tags and data.tag_sets then
error("Both `.tags` and `.tag_sets` cannot be specified")
end
end
local cur_infl = {}
local tag_sets = data.tag_sets
if not tag_sets then
tag_sets = export.split_tag_set(data.tags)
for i, tag_set in ipairs(tag_sets) do
tag_sets[i] = export.parse_tag_set_properties(tag_set)
end
end
 
local inflections = {}
local inflections = {}
local categories = {}
for _, tag_set in ipairs(tag_sets) do
local normalized_tag_sets = export.normalize_tag_set(tag_set.tags, data.lang, "do-track")


local ntags = export.normalize_tags(data.tags, nil, "do-track")
for _, normalized_tag_set in ipairs(normalized_tag_sets) do
 
local cur_infl = {}
for i, tagspec in ipairs(ntags) do
local this_categories, this_labels = export.fetch_categories_and_labels(normalized_tag_set, data.lang,
if tagspec == ";" then
data.POS, data.pagename, type(data.lemmas) == "table" and data.lemmas or nil)
if #cur_infl > 0 then
if not data.nocat then
table.insert(inflections, table.concat(cur_infl))
m_table.extendList(categories, this_categories)
end
end
 
local cur_infl = export.get_tag_set_display_form(normalized_tag_set, data.lang, data.joiner)
cur_infl = {}
else
local to_insert = export.get_tag_display_form(tagspec, data.joiner)
-- Maybe insert a space before inserting the display form
-- of the tag. We insert a space if
-- (a) we're not the first tag; and
-- (b) the tag we're about to insert doesn't have the
--    "no_space_on_left" property; and
-- (c) the preceding tag doesn't have the "no_space_on_right"
--    property.
-- NOTE: We depend here on the fact that
-- (1) all tags with either of the above properties set have the
--    same display form as canonical form, and
-- (2) all tags with either of the above properties set are
--    single-character tags.
-- The second property is an optimization to avoid looking up
-- display forms resulting from multipart tags, which won't be
-- found and which will trigger loading of [[Module:form of/data2]].
-- If multichar punctuation is added in the future, it's ok to
-- change the == 1 below to <= 2 or <= 3.
--
-- If the first property above fails to hold in the future, we
-- need to track the canonical form of each tag (including the
-- previous one) as well as the display form. This would also
-- avoid the need for the == 1 check.
if #cur_infl > 0 then
if #cur_infl > 0 then
local most_recent_tagobj = ulen(cur_infl[#cur_infl]) == 1 and
if tag_set.labels then
export.lookup_tag(cur_infl[#cur_infl])
this_labels = m_table.append(tag_set.labels, this_labels)
local to_insert_tagobj = ulen(to_insert) == 1 and
export.lookup_tag(to_insert)
if (
(not most_recent_tagobj or
not most_recent_tagobj.no_space_on_right) and
(not to_insert_tagobj or
not to_insert_tagobj.no_space_on_left)
) then
table.insert(cur_infl, " ")
end
end
table.insert(inflections, {infl_text = cur_infl, labels = this_labels})
end
end
table.insert(cur_infl, to_insert)
end
end
end
end


if #cur_infl > 0 then
local overall_labels, need_per_tag_set_labels
table.insert(inflections, table.concat(cur_infl))
for _, inflection in ipairs(inflections) do
if overall_labels == nil then
overall_labels = inflection.labels
elseif not m_table.deepEquals(overall_labels, inflection.labels) then
need_per_tag_set_labels = true
overall_labels = nil
break
end
end
end


local format_data = require("Module:table").shallowcopy(data)
if not need_per_tag_set_labels then
for _, inflection in ipairs(inflections) do
inflection.labels = nil
end
end
 
local format_data = m_table.shallowcopy(data)
 
local function format_labels(labels, notext)
if labels and #labels > 0 then
return require(labels_module).show_labels { labels = labels, lang = data.lang, sort = data.sort, nocat = data.nocat } ..
(notext and (data.pretext or "") == "" and "" or " ")
else
return ""
end
end


local of_text = data.lemmas and " of" or ""
local formatted_text
if #inflections == 1 then
if #inflections == 1 then
format_data.text =
if need_per_tag_set_labels then
data.notext and "" or ((data.capfirst and require("Module:string utilities").ucfirst(inflections[1]) or inflections[1]) ..
error("Internal error: need_per_tag_set_labels should not be set with one inflection")
(data.terminfo and " of" or ""))
end
return export.format_form_of(format_data)
format_data.text = format_labels(overall_labels, data.notext) .. (data.pretext or "") .. (data.notext and "" or
((data.capfirst and require("Module:string utilities").ucfirst(inflections[1].infl_text) or inflections[1].infl_text) .. of_text))
formatted_text = export.format_form_of(format_data)
else
else
format_data.text = data.notext and "" or ((data.capfirst and "Inflection" or "inflection") ..
format_data.text = format_labels(overall_labels, data.notext) .. (data.pretext or "") .. (data.notext and "" or
(data.terminfo and " of" or ""))
((data.capfirst and "Inflection" or "inflection") .. of_text))
format_data.posttext = (data.posttext or "") .. ":"
format_data.posttext = (data.posttext or "") .. ":"
local link = export.format_form_of(format_data)
local link = export.format_form_of(format_data)
local text_classes = data.text_classes or "form-of-definition use-with-mention"
local text_classes = data.text_classes or "form-of-definition use-with-mention"
return link .."\n## <span class='" .. text_classes .. "'>" ..
for i, inflection in ipairs(inflections) do
table.concat(inflections, "</span>\n## <span class='" .. text_classes .. "'>") .. "</span>"
inflections[i] = "\n## " .. format_labels(inflection.labels, false) ..
"<span class='" .. text_classes .. "'>" .. inflection.infl_text .. "</span>"
end
formatted_text = link .. table.concat(inflections)
end
end
if not data.no_format_categories then
if #categories > 0 then
formatted_text = formatted_text .. require("Module:utilities").format_categories(categories, data.lang,
data.sort, nil, export.force_cat)
end
return formatted_text
end
return formatted_text, categories
end
end


function export.to_Wikidata_IDs(tags, skip_tags_without_ids)
if type(tags) == "string" then
tags = mw.text.split(tags, "|", true)
end


--[==[
Given a tag set, return a flattened list all Wikidata ID's of all tags in the tag set. FIXME: Only used in a debugging
function in [[Module:se-verbs]]; move there.
]==]
function export.to_Wikidata_IDs(tag_set, lang, skip_tags_without_ids)
local ret = {}
local ret = {}


local function get_wikidata_id(tag)
local function get_wikidata_id(tag)
if tag == ";" and not skip_tags_without_ids then
local data = export.lookup_tag(tag, lang)
error("Semicolon is not supported for Wikidata IDs")
else
return nil
end
 
local data = export.lookup_tag(tag)


if not data or not data.wikidata then
if not data or not data[export.WIKIDATA] then
if not skip_tags_without_ids then
if not skip_tags_without_ids then
error("The tag \"" .. tag .. "\" does not have a Wikidata ID defined in [[Module:form of/data]]")
error('The tag "' .. tag .. '" does not have a Wikidata ID defined in the form-of data modules')
else
else
return nil
return nil
end
end
else
else
return data.wikidata
return ("Q%s"):format(data[export.WIKIDATA])
end
end
end
end


for i, tag in ipairs(export.normalize_tags(tags)) do
local normalized_tag_sets = export.normalize_tag_set(tag_set, lang)
if type(tag) == "table" then
for _, tag_set in ipairs(normalized_tag_sets) do
local ids = {}
for _, tag in ipairs(tag_set) do
for _, onetag in ipairs(tag) do
if type(tag) == "table" then
table.insert(ids, get_wikidata_id(onetag))
for _, subtag in ipairs(tag) do
if type(subtag) == "table" then
-- two-level multipart tag; FIXME: delete support for this
for _, subsubtag in ipairs(subtag) do
table.insert(ret, get_wikidata_id(subsubtag))
end
else
table.insert(ret, get_wikidata_id(subtag))
end
end
else
table.insert(ret, get_wikidata_id(tag))
end
end
table.insert(ret, ids)
else
table.insert(ret, get_wikidata_id(tag))
end
end
end
end
Line 897: Line 1,353:
function export.dump_form_of_data(frame)
function export.dump_form_of_data(frame)
local data = {
local data = {
data = require("Module:form of/data"),
data = require(export.form_of_data_module),
data2 = require("Module:form of/data2")
data2 = require(export.form_of_data2_module)
}
}
return require("Module:JSON").toJSON(data)
return require("Module:JSON").toJSON(data)
end
function export.finalize_tag_data(tags, shortcuts)
local function process_shortcut(name, shortcut)
-- If the shortcut is already in the list, then there is a duplicate.
if shortcuts[shortcut] then
error("The shortcut \"" .. shortcut .. "\" (for the inflection tag \"" .. name .. "\") conflicts with an existing shortcut for the tag \"" .. shortcuts[shortcut] .. "\".")
elseif tags[shortcut] then
error("The shortcut \"" .. shortcut .. "\" (for the inflection tag \"" .. name .. "\") conflicts with an existing tag with that name.")
end
shortcuts[shortcut] = name
end
for name, data in pairs(tags) do
local data_shortcuts = data[export.SHORTCUTS]
if data_shortcuts then
if type(data_shortcuts) == "string" then
process_shortcut(name, data_shortcuts)
else
for _, shortcut in ipairs(data_shortcuts) do
process_shortcut(name, shortcut)
end
end
end
end
end
end




return export
return export

Latest revision as of 18:12, 12 August 2024



local export = {}

export.force_cat = false -- for testing; set to true to display categories even on non-mainspace pages

local m_links = require("Module:links")
local m_string_utils = require("Module:string utilities")
local m_table = require("Module:table")
local parse_utilities_module = "Module:parse utilities"
local labels_module = "Module:labels"
local utilities_module = "Module:utilities"
export.form_of_pos_module = "Module:form of/pos"
export.form_of_functions_module = "Module:form of/functions"
export.form_of_cats_module = "Module:form of/cats"
export.form_of_lang_data_module_prefix = "Module:form of/lang-data/"
export.form_of_data_module = "Module:form of/data"
export.form_of_data2_module = "Module:form of/data2"

local ulen = m_string_utils.len
local rsubn = m_string_utils.gsub
local rmatch = m_string_utils.match
local rsplit = m_string_utils.split

export.TAG_TYPE = 1
export.GLOSSARY = 2
export.SHORTCUTS = 3
export.WIKIDATA = 4

export.APPENDIX = true
export.WP = false
export.WIKT = 0

--[==[
Set listing the languages with lang-specific tags. If a language isn't listed here, the tags for that language won't be
recognized.
]==]
export.langs_with_lang_specific_tags = {
	["en"] = true,
	["got"] = true,
	["lt"] = true,
	["lv"] = true,
	["nl"] = true,
	["pi"] = true,
	["sw"] = true,
	["ttj"] = true,
}

--[==[ intro:

This module implements the underlying processing of {{tl|form of}}, {{tl|inflection of}} and specific variants such as
{{tl|past participle of}} and {{tl|alternative spelling of}}. Most of the logic in this file is to handle tags in
{{tl|inflection of}}. Other related files:

* [[Module:form of/templates]] contains the majority of the logic that implements the templates themselves.
* [[Module:form of/data]] is a data-only file containing information on the more common inflection tags, listing the
  tags, their shortcuts, the category they belong to (tense-aspect, case, gender, voice-valence, etc.), the appropriate
  glossary link and the wikidata ID.
* [[Module:form of/data2]] is a data-only file containing information on the less common inflection tags, in the same
  format as [[Module:form of/data]].
* [[Module:form of/lang-data/LANGCODE]] is a data-only file containing information on the language-specific inflection
  tags for the language with code LANGCODE, in the same format as [[Module:form of/data]]. Language-specific tags
  override general tags.
* [[Module:form of/cats]] is a data-only file listing the language-specific categories that are added when the
  appropriate combinations of tags are seen for a given language.
* [[Module:form of/pos]] is a data-only file listing the recognized parts of speech and their abbreviations, used for
  categorization. FIXME: This should be unified with the parts of speech listed in [[Module:links]].
* [[Module:form of/functions]] contains functions for use with [[Module:form of/data]] and [[Module:form of/cats]].
  They are contained in this module because data-only modules can't contain code. The functions in this file are of two
  types:
*# Display handlers allow for customization of the display of multipart tags (see below). Currently there is only
   one such handler, for handling multipart person tags such as `1//2//3`.
*# Cat functions allow for more complex categorization logic, and are referred to by name in [[Module:form of/cats]].
   Currently no such functions exist.

The following terminology is used in conjunction with {{tl|inflection of}}:

* A ''tag'' is a single grammatical item, as specified in a single numbered parameter of {{tl|inflection of}}. Examples
  are `masculine`, `nominative`, or `first-person`. Tags may be abbreviated, e.g. `m` for `masculine`, `nom` for
  `nominative`, or `1` for `first-person`. Such abbreviations are called ''aliases'', and some tags have multiple
  equivalent aliases (e.g. `p` or `pl` for `plural`). The full, non-abbreviated form of a tag is called its
  ''canonical form''.
* The ''display form'' of a tag is the way it's displayed to the user. Usually the displayed text of the tag is the same
  as its canonical form, and it normally functions as a link to a glossary entry explaining the tag. Usually the link is
  to an entry in [[Appendix:Glossary]], but sometimes the tag is linked to an individual dictionary entry or to a
  Wikipedia entry. Occasionally, the display text differs from the canonical form of the tag. An example is the tag
  `comparative case`, which has the display text read as simply `comparative`. Normally, tags referring to cases don't
  have the word "case" in them, but in this case the tag `comparative` was already used as an alias for the tag
  `comparative degree`, so the tag was named `comparative case` to avoid clashing. A similar situation occurs with
  `adverbial case` vs. the grammar tag `adverbial` (as in `adverbial participle`).
* A ''tag set'' is an ordered list of tags, which together express a single inflection, for example, `1|s|pres|ind`,
  which can be expanded to canonical-form tags as `first-person|singular|present|indicative`.
* A ''conjoined tag set'' is a tag set that consists of multiple individual tag sets separated by a semicolon, e.g.
  `1|s|pres|ind|;|2|s|imp`, which specifies two tag sets, `1|s|pres|ind` as above and `2|s|imp` (in canonical form,
  `second-person|singular|imperative`). Multiple tag sets specified in a single call to {{tl|inflection of}} are
  specified in this fashion. Conjoined tag sets can also occur in list-tag shortcuts.
* A ''multipart tag'' is a tag that embeds multiple tags within it, such as `f//n` or `nom//acc//voc`. These are used in
  the case of [[syncretism]], when the same form applies to multiple inflections. Examples are the Spanish present
  subjunctive, where the first-person and third-person singular have the same form (e.g. {{m|es|siga}} from
  {{m|es|seguir|t=to follow}}), or Latin third-declension adjectives, where the dative and ablative plural of all
  genders have the same form (e.g. {{m|la|omnibus}} from {{m|la|omnis|t=all}}). These would be expressed respectively as
  `1//3|s|pres|sub` and `dat//abl|m//f//n|p`, where the use of the multipart tag compactly encodes the syncretism and
  avoids the need to individually list out all of the inflections. Multipart tags currently display as a list separated
  by a slash, e.g.  ''dative/ablative'' or ''masculine/feminine/neuter'' where each individual word is linked
  appropriately. As a special case, multipart tags involving persons display specially; for example, the multipart tag
  `1//2//3` displays as ''first-, second- and third-person'', with the word "person" occurring only once.
* A ''two-level multipart tag'' is a special type of multipart tag that joins two or more tag sets instead of joining
  individual tags. The tags within the tag set are joined by a colon, e.g. `1:s//3:p`, which is displayed as
  ''first-person singular and third-person plural'', e.g. for use with the form {{m|grc|μέλλον}} of the verb
  {{m|grc|μέλλω|t=to intend}}, which uses the tag set `1:s//3:p|impf|actv|indc|unaugmented` to express the syncretism
  between the first singular and third plural forms of the imperfect active indicative unaugmented conjugation.
  Two-level multipart tags should be used sparingly; if in doubt, list out the inflections separately. [FIXME: Make
  two-level multipart tags obsolete.]
* A ''shortcut'' is a tag that expands to any type of tag described above, or to any type of tag set described above.
  Aliases are a particular type of shortcut whose expansion is a single non-multipart tag.
* A ''multipart shortcut'' is a shortcut that expands into a multipart tag, for example `123`, which expands to the
  multipart tag `1//2//3`. Only the most common such combinations exist as shortcuts.
* A ''list shortcut'' is a special type of shortcut that expands to a list of tags instead of a single tag. For example,
  the shortcut `1s` expands to `1|s` (first-person singular). Only the most common such combinations exist as shortcuts.
* A ''conjoined shortcut'' is a special type of list shortcut that consists of a conjoined tag set (multiple logical tag
  sets). For example, the English language-specific shortcut `ed-form` expands to `spast|;|past|part`, expressing the
  common syncretism between simple past and past participle in English (and in this case, `spast` is itself a list
  shortcut that expands to `simple|past`).
]==]

-- version of rsubn() that discards all but the first return value
local function rsub(term, foo, bar)
	local retval = rsubn(term, foo, bar)
	return retval
end


local function normalize_index(list, index)
	if index < 0 then
		return #list + index + 1
	end
	return index
end

-- FIXME, consider moving to [[Module:table]]
-- Return true if the list `tags1`, treated as a set, is a subset of the list `tags2`, also treated as a set.
local function is_subset(tags1, tags2)
	tags1 = m_table.listToSet(tags1)
	tags2 = m_table.listToSet(tags2)
	for tag, _ in pairs(tags1) do
		if not tags2[tag] then
			return false
		end
	end
	return true
end


-- FIXME, move to [[Module:table]]
local function slice(list, i, j)
	--checkType("slice", 1, list, "table")
	--checkType("slice", 2, i, "number", true)
	--checkType("slice", 3, j, "number", true)
	if i == nil then
		i = 1
	else
		i = normalize_index(list, i)
	end
	j = normalize_index(list, j or -1)

	local retval = {}
	local k = 0
	for index = i, j do
		k = k + 1
		retval[k] = list[index]
	end
	return retval
end


local function wrap_in_span(text, classes)
	return ("<span class='%s'>%s</span>"):format(classes, text)
end


--[==[
Lowest-level implementation of form-of templates, including the general {{tl|form of}} as well as those that deal with
inflection tags, such as the general {{tl|inflection of}}, semi-specific variants such as {{tl|participle of}}, and
specific variants such as {{tl|past participle of}}. `data` contains all the information controlling the display, with
the following fields:

* `.text`: Text to insert before the lemmas. Wrapped in the value of `.text_classes`, or its default; see below.
* `.lemmas`: List of objects describing the lemma(s) of which the term in question is a non-lemma form. These are passed
   directly to {full_link()} in [[Module:links]]. Each object should have at minimum a `.lang` field containing the
   language of the lemma and a `.term` field containing the lemma itself. Each object is formatted using {full_link()}
   and then if there are more than one, they are joined using {serialCommaJoin()} in [[Module:table]]. Alternatively,
   `.lemmas` can be a string, which is displayed directly, or omitted, to show no lemma links and omit the connecting
   text.
* `.lemma_face`: "Face" to use when displaying the lemma objects. Usually should be set to {"term"}.
* `.enclitics`: List of enclitics to display after the lemmas, in parens.
* `.base_lemmas`: List of base lemmas to display after the lemmas, in the case where the lemmas in `.lemmas` are
   themselves forms of another lemma (the base lemma), e.g. a comparative, superlative or participle. Each object is of
   the form { { paramobj = PARAM_OBJ, lemmas = {LEMMA_OBJ, LEMMA_OBJ, ...} }} where PARAM_OBJ describes the properties
   of the base lemma parameter (i.e. the relationship between the intermediate and base lemmas) and LEMMA_OBJ is an
   object suitable to be passed to {full_link()} in [[Module:links]]. PARAM_OBJ is of the format
   { { param = "PARAM", tags = {"TAG", "TAG", ...} } where PARAM is the name of the parameter to {{tl|inflection of}}
   etc. that holds the base lemma(s) of the specified relationship and the tags describe the relationship, such as
   { {"comd"}} or { {"past", "part"}}.
* `.text_classes`: CSS classes used to wrap the tag text and lemma links. Default is {"form-of-definition use-with-mention"}
   for the tag text and lemma links, and additionally {"form-of-definition-link"} specifically for the lemma links.
   (FIXME: Should separate out the lemma links into their own field.)
* `.posttext`: Additional text to display after the lemma links.
]==]
function export.format_form_of(data)
	if type(data) ~= "table" then
		error("Internal error: First argument must now be a table of arguments")
	end
	local text_classes = data.text_classes or "form-of-definition use-with-mention"
	local lemma_classes = data.text_classes or "form-of-definition-link"
	local parts = {}
	local function ins(text)
		table.insert(parts, text)
	end
	ins("<span class='" .. text_classes .. "'>")
	ins(data.text)
	if data.text ~= "" and data.lemmas then
		ins(" ")
	end
	if data.lemmas then
		if type(data.lemmas) == "string" then
			ins(wrap_in_span(data.lemmas, lemma_classes))
		else
			local formatted_terms = {}
			for _, lemma in ipairs(data.lemmas) do
				table.insert(formatted_terms, wrap_in_span(
					m_links.full_link(lemma, data.lemma_face), lemma_classes
				))
			end
			ins(m_table.serialCommaJoin(formatted_terms))
		end
	end
	if data.enclitics and #data.enclitics > 0 then
		-- The outer parens need to be outside of the text_classes span so they show in upright instead of italic, or
		-- they will clash with upright parens generated by link annotations such as transliterations and pos=.
		ins("</span>")
		local formatted_terms = {}
		for _, enclitic in ipairs(data.enclitics) do
			-- FIXME, should we have separate clitic face and/or classes?
			table.insert(formatted_terms, wrap_in_span(
				m_links.full_link(enclitic, data.lemma_face, nil, "show qualifiers"), lemma_classes
			))
		end
		ins(" (")
		ins(wrap_in_span("with enclitic" .. (#data.enclitics > 1 and "s" or "") .. " ", text_classes))
		ins(m_table.serialCommaJoin(formatted_terms))
		ins(")")
		ins("<span class='" .. text_classes .. "'>")
	end
	if data.base_lemmas and #data.base_lemmas > 0 then
		for _, base_lemma in ipairs(data.base_lemmas) do
			ins(", the </span>")
			ins(export.tagged_inflections {
				lang = base_lemma.lemmas[1].lang,
				tags = base_lemma.paramobj.tags,
				lemmas = base_lemma.lemmas,
				lemma_face = data.lemma_face,
				no_format_categories = true,
				nocat = true,
				text_classes = data.text_classes,
			})
			ins("<span class='" .. text_classes .. "'>")
		end
	end
	-- FIXME, should posttext go before enclitics? If so we need to have separate handling for the
	-- final colon when there are multiple tag sets in tagged_inflections().
	if data.posttext then
		ins(data.posttext)
	end
	ins("</span>")
	return table.concat(parts)
end


--[==[
Return true if `tag` contains an internal link or HTML.
]==]
function export.is_link_or_html(tag)
	return tag:find("[[", nil, true) or tag:find("|", nil, true) or tag:find("<", nil, true)
end


--[==[
Look up a tag (either a shortcut of any sort of a canonical long-form tag) and return its expansion. The expansion
will be a string unless the shortcut is a list-tag shortcut such as `1s`; in that case, the expansion will be a
list. The caller must handle both cases. Only one level of expansion happens; hence, `acc` expands to {"accusative"},
`1s` expands to { {"1", "s"}} (not to { {"first", "singular"}}) and `123` expands to {"1//2//3"}. The expansion will be
the same as the passed-in tag in the following circumstances:

# The tag is `;` (this is special-cased, and no lookup is done).
# The tag is a multipart tag such as `nom//acc` (this is special-cased, and no lookup is done).
# The tag contains a raw link (this is special-cased, and no lookup is done).
# The tag contains HTML (this is special-cased, and no lookup is done).
# The tag is already a canonical long-form tag.
# The tag is unrecognized.

This function first looks up in the lang-specific data module [[Module:form of/lang-data/LANGCODE]], then in
[[Module:form of/data]] (which includes more common non-lang-specific tags) and finally (only if the tag is not
recognized as a shortcut or canonical tag, and is not of types 1-4 above) in [[Module:form of/data2]].

If the expansion is a string and is different from the tag, track it if `do_track` is true.
]==]
function export.lookup_shortcut(tag, lang, do_track)
	-- If there is HTML or a link in the tag, return it directly; don't try
	-- to look it up, which will fail.
	if tag == ";" or tag:find("//", nil, true) or export.is_link_or_html(tag) then
		return tag
	end
	local expansion
	local langcode = lang and lang:getCode()
	if langcode and export.langs_with_lang_specific_tags[langcode] then
		local langdata = mw.loadData(export.form_of_lang_data_module_prefix .. langcode)
		-- If this is a canonical long-form tag, just return it, and don't check for shortcuts. This is an
		-- optimization; see below.
		if langdata.tags[tag] then
			return tag
		end
		expansion = langdata.shortcuts[tag]
	end
	if not expansion and lang then
		-- If the lang we're dealing with is an etym-only lang, try again with the corresponding full language.
		local full_langcode = lang:getFullCode()
		if full_langcode ~= langcode and export.langs_with_lang_specific_tags[full_langcode] then
			local langdata = mw.loadData(export.form_of_lang_data_module_prefix .. full_langcode)
			-- If this is a canonical long-form tag, just return it, and don't check for shortcuts. This is an
			-- optimization; see below.
			if langdata.tags[tag] then
				return tag
			end
			expansion = langdata.shortcuts[tag]
		end
	end
	if not expansion then
		local m_data = mw.loadData(export.form_of_data_module)
		-- If this is a canonical long-form tag, just return it, and don't check for shortcuts (which will cause
		-- [[Module:form of/data2]] to be loaded, because there won't be a shortcut entry in [[Module:form of/data]] --
		-- or, for that matter, in [[Module:form of/data2]]). This is an optimization; the code will still work without
		-- it, but will use up more memory.
		if m_data.tags[tag] then
			return tag
		end
		expansion = m_data.shortcuts[tag]
	end
	if not expansion then
		local m_data2 = mw.loadData(export.form_of_data2_module)
		expansion = m_data2.shortcuts[tag]
	end
	if not expansion then
		return tag
	end
	
	return expansion
end


--[==[
Look up a normalized/canonicalized tag and return the data object associated with it. If the tag isn't found, return
nil. This first looks up in the lang-specific data module [[Module:form of/lang-data/LANGCODE]], then in
[[Module:form of/data]] (which includes more common non-lang-specific tags) and then finally in
[[Module:form of/data2]].
]==]
function export.lookup_tag(tag, lang)
	local langcode = lang and lang:getCode()
	if langcode and export.langs_with_lang_specific_tags[langcode] then
		local langdata = mw.loadData(export.form_of_lang_data_module_prefix .. langcode)
		if langdata.tags[tag] then
			return langdata.tags[tag]
		end
	end
	local full_langcode = lang and lang:getFullCode()
	if full_langcode and full_langcode ~= langcode and export.langs_with_lang_specific_tags[full_langcode] then
		-- If the lang we're dealing with is an etym-only lang, try again with the corresponding full language.
		local langdata = mw.loadData(export.form_of_lang_data_module_prefix .. full_langcode)
		if langdata.tags[tag] then
			return langdata.tags[tag]
		end
	end
	local m_data = mw.loadData(export.form_of_data_module)
	local tagobj = m_data.tags[tag]
	if tagobj then
		return tagobj
	end
	local m_data2 = mw.loadData(export.form_of_data2_module)
	local tagobj2 = m_data2.tags[tag]
	if tagobj2 then
		return tagobj2
	end
	return nil
end


-- Normalize a single tag, which may be a shortcut but should not be a multipart tag, a multipart shortcut or a list
-- shortcut.
local function normalize_single_tag(tag, lang, do_track)
	local expansion = export.lookup_shortcut(tag, lang, do_track)
	if type(expansion) ~= "string" then
		error("Tag '" .. tag .. "' is a list shortcut, which is not allowed here")
	end
	tag = expansion
	return tag
end


--[=[
Normalize a component of a multipart tag. This should not have any // in it, but may join multiple individual tags with
a colon, and may be a single list-tag shortcut, which is treated as if colon-separated. The return value may be a list
of tags.
]=]
local function normalize_multipart_component(tag, lang, do_track)
	-- If there is HTML or a link in the tag, don't try to split on colon. A colon may legitimately occur in either one,
	-- and we don't want these things parsed. Note that we don't do this check before splitting on //, which we don't
	-- expect to occur in links or HTML; see comment in normalize_tag().
	if export.is_link_or_html(tag) then
		return tag
	end
	local components = rsplit(tag, ":", true)
	if #components == 1 then
		-- We allow list-tag shortcuts inside of multipart tags, e.g.
		-- '1s//3p'. Check for this now.
		tag = export.lookup_shortcut(tag, lang, do_track)
		if type(tag) == "table" then
			-- We found a list-tag shortcut; treat as if colon-separated.
			components = tag
		else
			return normalize_single_tag(tag, lang, do_track)
		end
	end
	local normtags = {}
	for _, component in ipairs(components) do
		table.insert(normtags, normalize_single_tag(component, lang, do_track))
	end

	return normtags
end


--[=[
Normalize a single tag. The return value may be a list (in the case of multipart tags), which will contain nested lists
in the case of two-level multipart tags.
]=]
local function normalize_tag(tag, lang, do_track)
	-- We don't check for links or HTML before splitting on //, which we don't expect to occur in links or HTML. Doing
	-- it this way allows for a tag like '{{lb|grc|Epic}}//{{lb|grc|Ionic}}' to function correctly (the template calls
	-- will be expanded before we process the tag, and will contain links and HTML). The only check we do is for a URL,
	-- which shouldn't normally occur, but might if the user tries to put an external link into the tag. URL's with //
	-- normally have the sequence ://, which should never normally occur when // and : are used in their normal ways.
	if tag:find("://", nil, true) then
		return tag
	end
	local split_tags = rsplit(tag, "//", true)
	if #split_tags == 1 then
		local retval = normalize_multipart_component(tag, lang, do_track)
		if type(retval) == "table" then
			-- The user gave a tag like '1:s', i.e. with colon but without //. Allow this, but we need to return a
			-- nested list.
			return {retval}
		end
		return retval
	end
	local normtags = {}
	for _, single_tag in ipairs(split_tags) do
		table.insert(normtags, normalize_multipart_component(single_tag, lang, do_track))
	end
	return normtags
end


--[==[
Normalize a tag set (a list of tags) into its canonical-form tags. The return value is a list of normalized tag sets
(a list because of there may be conjoined shortcuts among the input tags). A normalized tag set is a list of tag
elements, where each element is either a string (the canonical form of a tag), a list of such strings (in the case of
multipart tags) or a list of lists of such strings (in the case of two-level multipart tags). For example, the multipart
tag `nom//acc//voc` will be represented in canonical form as { {"nominative", "accusative", "vocative"}}, and the
two-level multipart tag `1:s//3:p` will be represented as { {{"first-person", "singular"}, {"third-person", "plural"}}}.

Example 1:

{normalize_tag_set({"nom//acc//voc", "n", "p"})} = { {{{"nominative", "accusative", "vocative"}, "masculine", "plural"}}}

Example 2:

{normalize_tag_set({"ed-form"}, ENGLISH)} = { {{"simple", "past"}, {"past", "participle"}}}

Example 3:

{normalize_tag_set({"archaic", "ed-form"}, ENGLISH)} = { {{"archaic", "simple", "past"}, {"archaic", "past", "participle"}}}
]==]
function export.normalize_tag_set(tag_set, lang, do_track)
	
	local output_tag_set = {}
	local saw_semicolon = false

	for _, tag in ipairs(tag_set) do
		-- Expand the tag, which may generate a new tag (either a fully canonicalized tag, a multipart tag, or a list
		-- of tags).
		tag = export.lookup_shortcut(tag, lang, do_track)
		if type(tag) == "table" then
			saw_semicolon = m_table.contains(tag, ";")
			if saw_semicolon then
				-- If we saw a conjoined shortcut, we need to use a more general algorithm that can expand a single
				-- tag set into multiple.
				break
			end

			for _, t in ipairs(tag) do
				table.insert(output_tag_set, normalize_tag(t, lang, do_track))
			end
		else
			table.insert(output_tag_set, normalize_tag(tag, lang, do_track))
		end
	end

	if not saw_semicolon then
		return {output_tag_set}
	end

	-- Use a more general algorithm that handles conjoined shortcuts.
	local output_tag_set = {}
	for i, tag in ipairs(tag_set) do
		-- Expand the tag, which may generate a new tag (either a fully canonicalized tag, a multipart tag, or a list
		-- of tags).
		tag = export.lookup_shortcut(tag, lang, do_track)
		if type(tag) == "table" then
			local output_tag_sets = {}
			local shortcut_tag_sets = export.split_tag_set(tag)
			local normalized_shortcut_tag_sets = {}
			for _, shortcut_tag_set in ipairs(shortcut_tag_sets) do
				m_table.extendList(normalized_shortcut_tag_sets,
					export.normalize_tag_set(shortcut_tag_set, lang, do_track))
			end
			local after_tags = slice(tag_set, i + 1)
			local normalized_after_tags_sets = export.normalize_tag_set(after_tags, lang, do_track)
			for _, normalized_shortcut_tag_set in ipairs(normalized_shortcut_tag_sets) do
				for _, normalized_after_tags_set in ipairs(normalized_after_tags_sets) do
					table.insert(output_tag_sets, m_table.append(output_tag_set, normalized_shortcut_tag_set,
						normalized_after_tags_set))
				end
			end
			return output_tag_sets
		else
			table.insert(output_tag_set, normalize_tag(tag, lang, do_track))
		end
	end

	error("Internal error: Should not get here")
end


function export.combine_multipart_tags(tag_set)
	for i, tag in ipairs(tag_set) do
		if type(tag) == "table" then
			for j, subtag in ipairs(tag) do
				if type(subtag) == "table" then
					tag[j] = table.concat(subtag, ":")
				end
			end
			tag_set[i] = table.concat(tag, "//")
		end
	end

	return tag_set
end


function export.normalize_tags(tags, lang, recombine_multitags, do_track)
	local tag_sets = export.normalize_tag_set(tags, lang, do_track)
	if recombine_multitags then
		for i, tag_set in ipairs(tag_sets) do
			tag_sets[i] = export.combine_multipart_tags(tag_set)
		end
		return export.combine_tag_sets(tag_sets)
	end
	return tag_sets
end


--[==[
Split a tag set containing two-level multipart tags into one or more tag sets not containing such tags.
Single-level multipart tags are left alone. (If we need to, a slight modification of the following code
will also split single-level multipart tags.) This assumes that multipart tags are represented as lists
and two-level multipart tags are represented as lists of lists, as is output by {normalize_tag_set()}.
NOTE: We have to be careful to properly handle imbalanced two-level multipart tags such as
`def:s//p` (or the reverse, `s//def:p`).
]==]
function export.split_two_level_multipart_tag_set(tag_set)
	for i, tag in ipairs(tag_set) do
		if type(tag) == "table" then
			-- We saw a multipart tag. Check if any of the parts are two-level.
			local saw_two_level_tag = false
			for _, first_level_tag in ipairs(tag) do
				if type(first_level_tag) == "table" then
					saw_two_level_tag = true
					break
				end
			end
			if saw_two_level_tag then
				-- We found a two-level multipart tag.
				-- (1) Extract the preceding tags.
				local pre_tags = slice(tag_set, 1, i - 1)
				-- (2) Extract the following tags.
				local post_tags = slice(tag_set, i + 1)
				-- (3) Loop over each tag set alternant in the two-level multipart tag.
				-- For each alternant, form the tag set consisting of pre_tags + alternant + post_tags,
				-- and recursively split that tag set.
				local resulting_tag_sets = {}
				for _, first_level_tag_set in ipairs(tag) do
					local expanded_tag_set = {}
					m_table.extendList(expanded_tag_set, pre_tags)
					-- The second level may have a string or a list.
					if type(first_level_tag_set) == "table" then
						m_table.extendList(expanded_tag_set, first_level_tag_set)
					else
						table.insert(expanded_tag_set, first_level_tag_set)
					end
					m_table.extendList(expanded_tag_set, post_tags)
					m_table.extendList(resulting_tag_sets, export.split_two_level_multipart_tag_set(expanded_tag_set))
				end
				return resulting_tag_sets
			end
		end
	end

	return {tag_set}
end


--[==[
Split a tag set that may consist of multiple semicolon-separated tag sets into the component tag sets.
]==]
function export.split_tag_set(tag_set)
	local split_tag_sets = {}
	local cur_tag_set = {}
	for _, tag in ipairs(tag_set) do
		if tag == ";" then
			if #cur_tag_set > 0 then
				table.insert(split_tag_sets, cur_tag_set)
			end
			cur_tag_set = {}
		else
			table.insert(cur_tag_set, tag)
		end
	end
	if #cur_tag_set > 0 then
		table.insert(split_tag_sets, cur_tag_set)
	end
	return split_tag_sets
end

export.split_tags_into_tag_sets = export.split_tag_set


--[==[
Combine multiple tag sets in a tag set group into a simple tag set, with logical tag sets separated by semicolons.
This is the opposite of {split_tag_set()}.
]==]
function export.combine_tag_sets(tag_sets)
	if #tag_sets == 1 then
		return tag_sets[1]
	end
	local combined_tag_set = {}
	for _, tag_set in ipairs(tag_sets) do
		if #combined_tag_set > 0 then
			table.insert(combined_tag_set, ";")
		end
		m_table.extendList(combined_tag_set, tag_set)
	end
	return tags
end


local tag_set_param_mods = {
	lb = {
		item_dest = "labels",
		convert = function(arg, parse_err)
			return rsplit(arg, "//", true)
		end,
	}
}


--[==[
Parse tag set properties from a tag set (list of tags). Currently no per-tag properties are recognized, and the only
per-tag-set property recognized is `<lb:...>` for specifing label(s) for the tag set. Per-tag-set properties must be
attached to the last tag.
]==]
function export.parse_tag_set_properties(tag_set)
	local function generate_tag_set_obj(last_tag)
		tag_set[#tag_set] = last_tag
		return {tags = tag_set}
	end
	local last_tag = tag_set[#tag_set]
	-- Check for inline modifier, e.g. מרים<tr:Miryem>. But exclude HTML entry with <span ...>, <i ...>, <br/> or
	-- similar in it, caused by wrapping an argument in {{l|...}}, {{af|...}} or similar. Basically, all tags of
	-- the sort we parse here should consist of a less-than sign, plus letters, plus a colon, e.g. <lb:...>, so if
	-- we see a tag on the outer level that isn't in this format, we don't try to parse it. The restriction to the
	-- outer level is to allow generated HTML inside of e.g. qualifier tags, such as foo<q:similar to {{m|fr|bar}}>.
	if last_tag:find("<") and not last_tag:find("^[^<]*<[a-z]*[^a-z:]") then
		return require(parse_utilities_module).parse_inline_modifiers(last_tag, {
			param_mods = tag_set_param_mods,
			generate_obj = generate_tag_set_obj,
		})
	else
		return generate_tag_set_obj(last_tag)
	end
end


function export.normalize_pos(pos)
	if not pos then
		return nil
	end
	return mw.loadData(export.form_of_pos_module)[pos] or pos
end


-- Return the display form of a single canonical-form tag. The value
-- passed in must be a string (i.e. it cannot be a list describing a
-- multipart tag). To handle multipart tags, use get_tag_display_form().
local function get_single_tag_display_form(normtag, lang)
	local data = export.lookup_tag(normtag, lang)
	local display = normtag

	-- If the tag has a special display form, use it
	if data and data.display then
		display = data.display
	end

	-- If there is a nonempty glossary index, then show a link to it
	local glossary = data and data[export.GLOSSARY]
	if glossary ~= nil then
		if glossary == export.WIKT then
			display = "[[" .. normtag .. "|" .. display .. "]]"
		elseif glossary == export.WP then
			display = "[[w:" .. normtag .. "|" .. display .. "]]"
		elseif glossary == export.APPENDIX then
			display = "[[Appendix:Glossary#" .. mw.uri.anchorEncode(normtag) .. "|" .. display .. "]]"
		elseif type(glossary) ~= "string" then
			error(("Internal error: Wrong type %s for glossary value %s for tag %s"):format(
				type(glossary), mw.dumpObject(glossary), normtag))
		else
			local link = rmatch(glossary, "^wikt:(.*)")
			if link then
				display = "[[" .. link .. "|" .. display .. "]]"
			end
			if not link then
				link = rmatch(glossary, "^w:(.*)")
				if link then
					display = "[[w:" .. link .. "|" .. display .. "]]"
				end
			end
			if not link then
				display = "[[Appendix:Glossary#" .. mw.uri.anchorEncode(glossary) .. "|" .. display .. "]]"
			end
		end
	end
	return display
end


--[==[
Turn a canonicalized tag spec (which describes a single, possibly multipart tag) into the displayed form. The tag spec
may be a string (a canonical-form tag); a list of canonical-form tags (in the case of a simple multipart tag); or a
list of mixed canonical-form tags and lists of such tags (in the case of a two-level multipart tag). `joiner` indicates
how to join the parts of a multipart tag, and can be either {"and"} ("foo and bar", or "foo, bar and baz" for 3 or
more), {"slash"} ("foo/bar"), {"en-dash"} ("foo–bar") or {nil}, which uses the global default found in
{multipart_join_strategy()} in [[Module:form of/functions]]. (NOTE: The global default is {"slash"} and this seems
unlikely to change.)
]==]
function export.get_tag_display_form(tagspec, lang, joiner)
	if type(tagspec) == "string" then
		return get_single_tag_display_form(tagspec, lang)
	end
	-- We have a multipart tag. See if there's a display handler to display them specially.
	for _, handler in ipairs(require(export.form_of_functions_module).display_handlers) do
		local displayval = handler(tagspec, joiner)
		if displayval then
			return displayval
		end
	end
	-- No display handler.
	local displayed_tags = {}
	for _, first_level_tag in ipairs(tagspec) do
		if type(first_level_tag) == "string" then
			table.insert(displayed_tags, get_single_tag_display_form(first_level_tag, lang))
		else
			-- A first-level element of a two-level multipart tag. Currently we just separate the individual components
			-- with spaces, but other ways are possible, e.g. using an underscore, colon, parens or braces.
			local components = {}
			for _, component in ipairs(first_level_tag) do
				table.insert(components, get_single_tag_display_form(component, lang))
			end
			table.insert(displayed_tags, table.concat(components, " "))
		end
	end
	return require(export.form_of_functions_module).join_multiparts(displayed_tags, joiner)
end


--[==[
Given a normalized tag set (i.e. as output by {normalize_tag_set()}; all tags are in canonical form, multipart tags are
represented as lists, and two-level multipart tags as lists of lists), convert to displayed form (a string). See
{get_tag_display_form()} for the meaning of `joiner`.
]==]
function export.get_tag_set_display_form(normalized_tag_set, lang, joiner)
	local parts = {}

	for _, tagspec in ipairs(normalized_tag_set) do
		local to_insert = export.get_tag_display_form(tagspec, lang, joiner)
		-- Maybe insert a space before inserting the display form of the tag. We insert a space if
		-- (a) we're not the first tag; and
		-- (b) the tag we're about to insert doesn't have the "no_space_on_left" property; and
		-- (c) the preceding tag doesn't have the "no_space_on_right" property.
		-- NOTE: We depend here on the fact that
		-- (1) all tags with either of the above properties set have the same display form as canonical form, and
		-- (2) all tags with either of the above properties set are single-character tags.
		-- The second property is an optimization to avoid looking up display forms resulting from multipart tags,
		-- which won't be found and which will trigger loading of [[Module:form of/data2]]. If multichar punctuation is
		-- added in the future, it's ok to change the == 1 below to <= 2 or <= 3.
		--
		-- If the first property above fails to hold in the future, we need to track the canonical form of each tag
		-- (including the previous one) as well as the display form. This would also avoid the need for the == 1 check.
		if #parts > 0 then
			local most_recent_tagobj = ulen(parts[#parts]) == 1 and export.lookup_tag(parts[#parts], lang)
			local to_insert_tagobj = ulen(to_insert) == 1 and export.lookup_tag(to_insert, lang)
			if (
				(not most_recent_tagobj or not most_recent_tagobj.no_space_on_right) and
				(not to_insert_tagobj or not to_insert_tagobj.no_space_on_left)
			) then
				table.insert(parts, " ")
			end
		end
		table.insert(parts, to_insert)
	end

	return table.concat(parts)
end


--[==[
Given a normalized tag set (i.e. as output by {normalize_tag_set()}; all tags are in canonical form, multipart tags are
represented as lists, and two-level multipart tags as lists of lists), fetch the associated categories and labels.
Return two values, a list of categories and a list of labels. `lang` is the language of term represented by the tag set,
and `POS` is the user-provided part of speech (which may be {nil}).
]==]
function export.fetch_categories_and_labels(normalized_tag_set, lang, POS, pagename, lemmas)
	local m_cats = mw.loadData(export.form_of_cats_module)
	local categories = {}
	local labels = {}

	POS = export.normalize_pos(POS)
	-- First split any two-level multipart tags into multiple sets, to make our life easier.
	for _, tag_set in ipairs(export.split_two_level_multipart_tag_set(normalized_tag_set)) do
		-- Call a named function, either from the lang-specific data in
		-- [[Module:form of/lang-specific/LANGCODE/functions]] or in [[Module:form of/functions]].
		local function call_named_function(name, funtype)
			local data = {
				pagename = pagename or mw.title.getCurrentTitle().subpageText,
				lemmas = lemmas,
				tag_set = normalized_tag_set,
				lang = lang,
				POS = POS
			}
			local modules_tried = {}
			local function try_lang_specific_module(langcode)
				if export.langs_with_lang_specific_tags[langcode] then
					local lang_specific_module = export.form_of_lang_data_module_prefix .. langcode .. "/functions"
					local langdata = require(utilities_module).safe_require(lang_specific_module)
					if langdata then
						table.insert(modules_tried, lang_specific_module)
						if langdata.cat_functions then
							local fn = langdata.cat_functions[name]
							if fn then
								return fn(data), true
							end
						end
					end
				end
				return nil, false
			end
			-- First try lang-specific.
			local langcode = lang and lang:getCode()
			if langcode then
				local retval, found_it = try_lang_specific_module(langcode)
				if found_it then
					return retval
				end
			end
			-- If the lang we're dealing with is an etym-only lang, try again with the corresponding full language.
			local full_langcode = lang and lang:getFullCode()
			if full_langcode and full_langcode ~= langcode then
				local retval, found_it = try_lang_specific_module(full_langcode)
				if found_it then
					return retval
				end
			end
			-- Try lang-independent.
			table.insert(modules_tried, export.form_of_functions_module)
			local fn = require(export.form_of_functions_module).cat_functions[name]
			if fn then
				return fn(data)
			end
			for i, modname in ipairs(modules_tried) do
				modules_tried[i] = "[[" .. modname .. "]]"
			end
			error(("No %s function named '%s' in %s"):format(funtype, name, lang_specific_part,
				m_table.serialCommaJoin(modules_tried, {conj = "or", dontTag = true})))
		end

		-- Given a tag from the current tag set (which may be a list in case of a multipart tag),
		-- and a tag from a categorization spec, check that the two match.
		-- (1) If both are strings, we just check for equality.
		-- (2) If the spec tag is a string and the tag set tag is a list (i.e. it originates from a
		-- multipart tag), we check that the spec tag is in the list. This is because we want to treat
		-- multipart tags in user-specified tag sets as if the user had specified multiple tag sets.
		-- For example, if the user said "1//3|s|pres|ind" and the categorization spec says {"has", "1"},
		-- we want this to match, because "1//3|s|pres|ind" should be treated equivalently to two tag
		-- sets "1|s|pres|ind" and "3|s|pres|ind", and the former matches the categorization spec.
		-- (3) If the spec tag is a list (i.e. it originates from a multipart tag), we check that the
		-- tag set tag is also a list and is a superset of the spec tag. For example, if the categorization
		-- spec says {"has", "1//3"}, then the tag set tag must be a multipart tag that has both "1" and "3"
		-- in it. "1//3" works, as does "1//2//3".
		local function tag_set_tag_matches_spec_tag(tag_set_tag, spec_tag)
			if type(spec_tag) == "table" then
				if type(tag_set_tag) == "table" and is_subset(spec_tag, tag_set_tag) then
					return true
				end
			elseif type(tag_set_tag) == "table" then
				if m_table.contains(tag_set_tag, spec_tag) then
					return true
				end
			elseif tag_set_tag == spec_tag then
				return true
			end
			return false
		end

		-- Check that the current tag set matches the given spec tag. This means that any of the tags
		-- in the current tag set match, according to tag_set_tag_matches_spec_tag(); see above. If the
		-- current tag set contains only string tags (i.e. no multipart tags), and the spec tag is a
		-- string (i.e. not a multipart tag), this boils down to list containment, but it gets more
		-- complex when multipart tags are present.
		local function tag_set_matches_spec_tag(spec_tag)
			spec_tag = normalize_tag(spec_tag, lang)
			for _, tag_set_tag in ipairs(tag_set) do
				if tag_set_tag_matches_spec_tag(tag_set_tag, spec_tag) then
					return true
				end
			end
			return false
		end

		-- Check whether the given spec matches the current tag set. Two values are returned:
		-- (1) whether the spec matches the tag set; (2) the index of the category to add if
		-- the spec matches.
		local function check_condition(spec)
			if type(spec) == "boolean" then
				return spec
			elseif type(spec) ~= "table" then
				error("Wrong type of condition " .. spec .. ": " .. type(spec))
			end
			local predicate = spec[1]
			if predicate == "has" then
				return tag_set_matches_spec_tag(spec[2]), 3
			elseif predicate == "hasall" then
				for _, tag in ipairs(spec[2]) do
					if not tag_set_matches_spec_tag(tag) then
						return false, 3
					end
				end
				return true, 3
			elseif predicate == "hasany" then
				for _, tag in ipairs(spec[2]) do
					if tag_set_matches_spec_tag(tag) then
						return true, 3
					end
				end
				return false, 3
			elseif predicate == "tags=" then
				local normalized_spec_tag_sets = export.normalize_tag_set(spec[2], lang)
				if #normalized_spec_tag_sets > 1 then
					error("Internal error: No support for conjoined shortcuts in category/label specs in "
						.. "[[Module:form of/cats]] when processing spec tag set " .. table.concat(spec[2], "|"))
				end
				local normalized_spec_tag_set = normalized_spec_tag_sets[1]
				-- Check for and disallow two-level multipart tags in the specs. FIXME: Remove this when we remove
				-- support for two-level multipart tags.
				for _, tag in ipairs(normalized_spec_tag_set) do
					if type(tag) == "table" then
						for _, subtag in ipairs(tag) do
							if type(subtag) == "table" then
								error("Internal error: No support for two-level multipart tags in category/label specs"
									.. "[[Module:form of/cats]] when processing spec tag set "
									.. table.concat(spec[2], "|"))
							end
						end
					end
				end
				-- Allow tags to be in different orders, and multipart tags to be in different orders. To handle this,
				-- we first check that both tag set tags and spec tags have the same length. If so, we sort the
				-- multipart tags in the tag set tags and spec tags, and then check that all tags in the spec tags are
				-- in the tag set tags.
				if #tag_set ~= #normalized_spec_tag_set then
					return false, 3
				end
				local tag_set_tags = m_table.deepcopy(tag_set)
				for i=1,#tag_set_tags do
					if type(tag_set_tags[i]) == "table" then
						table.sort(tag_set_tags[i])
					end
					if type(normalized_spec_tag_set[i]) == "table" then
						table.sort(normalized_spec_tag_set[i])
					end
				end
				for i=1,#tag_set_tags do
					if not m_table.contains(tag_set_tags, normalized_spec_tag_set[i]) then
						return false, 3
					end
				end
				return true, 3
			elseif predicate == "p=" then
				return POS == export.normalize_pos(spec[2]), 3
			elseif predicate == "pany" then
				for _, specpos in ipairs(spec[2]) do
					if POS == export.normalize_pos(specpos) then
						return true, 3
					end
				end
				return false, 3
			elseif predicate == "pexists" then
				return POS ~= nil, 2
			elseif predicate == "not" then
				local condval = check_condition(spec[2])
				return not condval, 3
			elseif predicate == "and" then
				local condval = check_condition(spec[2])
				if condval then
					condval = check_condition(spec[3])
				end
				return condval, 4
			elseif predicate == "or" then
				local condval = check_condition(spec[2])
				if not condval then
					condval = check_condition(spec[3])
				end
				return condval, 4
			elseif predicate == "call" then
				return fn(call_named_function(spec[2], "condition")), 3
			else
				error("Unrecognized predicate: " .. predicate)
			end
		end

		-- Process a given spec. This checks any conditions in the spec against the
		-- tag set, and insert any resulting categories into `categories`. Return value
		-- is true if the outermost condition evaluated to true and a category was inserted
		-- (this is used in {"cond" ...} conditions, which stop when a subcondition evaluates
		-- to true).
		local function process_spec(spec)
			if not spec then
				return false
			elseif type(spec) == "string" then
				-- A category. Substitute POS request with user-specified part of speech or default.
				spec = rsub(spec, "<<p=(.-)>>", function(default)
					return POS or export.normalize_pos(default)
				end)
				table.insert(categories, lang:getFullName() .. " " .. spec)
				return true
			elseif type(spec) == "table" and spec.labels then
				-- A label spec.
				for _, label in ipairs(spec.labels) do
					m_table.insertIfNot(labels, label)
				end
				return true
			elseif type(spec) ~= "table" then
				error("Wrong type of specification " .. spec .. ": " .. type(spec))
			end
			local predicate = spec[1]
			if predicate == "multi" then
				-- WARNING! #spec doesn't work for objects loaded from loadData()
				for i, sp in ipairs(spec) do
					if i > 1 then
						process_spec(sp)
					end
				end
				return true
			elseif predicate == "cond" then
				-- WARNING! #spec doesn't work for objects loaded from loadData()
				for i, sp in ipairs(spec) do
					if i > 1 and process_spec(sp) then
						return true
					end
				end
				return false
			elseif predicate == "call" then
				return process_spec(call_named_function(spec[2], "spec"))
			else
				local condval, ifspec = check_condition(spec)
				if condval then
					process_spec(spec[ifspec])
					return true
				else
					process_spec(spec[ifspec + 1])
					-- FIXME: Are we sure this is correct?
					return false
				end
			end
		end

		local langcode = lang:getCode()
		local langspecs = m_cats[langcode]
		if langspecs then
			for _, spec in ipairs(langspecs) do
				process_spec(spec)
			end
		end
		local full_code = lang:getFullCode()
		if full_code ~= langcode then
			local langspecs = m_cats[full_code]
			if langspecs then
				for _, spec in ipairs(langspecs) do
					process_spec(spec)
				end
			end
		end
		if full_code ~= "und" then
			local langspecs = m_cats["und"]
			if langspecs then
				for _, spec in ipairs(langspecs) do
					process_spec(spec)
				end
			end
		end
	end

	return categories, labels
end


--[==[
Implementation of templates that display inflection tags, such as the general {{tl|inflection of}}, semi-specific
variants such as {{tl|participle of}}, and specific variants such as {{tl|past participle of}}. `data` contains all the
information controlling the display, with the following fields:

* `.lang`: ('''''required''''') Language to use when looking up language-specific inflection tags, categories and
  labels, and for displaying categories and labels.
* `.tags`: ('''''required''' unless `.tag_sets` is given'') List of non-canonicalized inflection tags. Multiple tag sets
  can be indicated by a {";"} as one of the tags, and tag-set properties may be attached to the last tag of a tag set.
  The tags themselves may come directly from the user (as in {{tl|inflection of}}); come partly from the user (as in
  {{tl|participle of}}, which adds the tag `part` to user-specified inflection tags); or be entirely specified by the
  template (as in {{tl|past participle of}}).
* `.tag_sets`: ('''''required''' unless `.tags` is given'') List of non-canonicalized tag sets and associated
  per-tag-set properties. Each element of the list is an object of the form
  { {tags = {"TAG", "TAG", ...}, labels = {"LABEL", "LABEL", ...}}. If `.tag_sets` is specified, `.tags` should not be
  given and vice-versa. Specifying `.tag_sets` in place of tags allowed per-tag set labels to be specified; otherwise,
  there is no advantage. [[Module:pt-gl-inflections]] uses this functionality to supply labels like {"Brazil"} and
  {"Portugal"} associated with specific tag sets.
* `.lemmas`: ('''''recommended''''') List of objects describing the lemma(s) of which the term in question is a
  non-lemma form. These are passed directly to {full_link()} in [[Module:links]]. Each object should have at minimum a
  `.lang` field containing the language of the lemma and a `.term` field containing the lemma itself. Each object is
  formatted using {full_link()} and then if there are more than one, they are joined using {serialCommaJoin()} in
  [[Module:table]]. Alternatively, `.lemmas` can be a string, which is displayed directly. If omitted entirely, no lemma
  links are shown and the connecting "of" is also omitted.
* `.lemma_face`: ('''''recommended''''') "Face" to use when displaying the lemma objects. Usually should be set to
  {"term"}.
* `.POS`: ('''''recommended''''') Categorizing part-of-speech tag. Comes from the {{para|p}} or {{para|POS}} argument of
  {{tl|inflection of}}.
* `.pagename`: Page name of "current" page or nil to use the actual page title; for testing purposes.
* `.enclitics`: List of enclitics to display after the lemmas, in parens.
* `.no_format_categories`: If true, don't format the categories derived from the inflection tags; just return them.
* `.sort`: Sort key for formatted categories. Ignored when `.no_format_categories` = {true}.
* `.nocat`: Suppress computation of categories (even if `.no_format_categories` is not given).
* `.notext`: Disable display of all tag text and `inflection of` text. (FIXME: Maybe not implemented correctly.)
* `.capfirst`: Capitalize the first word displayed.
* `.pretext`: Additional text to display before the inflection tags, but after any top-level labels.
* `.posttext`: Additional text to display after the lemma links.
* `.text_classes`: CSS classes used to wrap the tag text and lemma links. Default is
  {"form-of-definition use-with-mention"} for the tag text, {"form-of-definition-link"} for the lemma links. (FIXME:
  Should separate out the lemma links into their own field.)
`.joiner`: Override the joiner (normally a slash) used to join multipart tags. You should normally not specify this.

A typical call might look like this (for {{m+|es|amo}}): {
	local lang = require("Module:languages").getByCode("es")

	local lemma_obj = {
		lang = lang,
		term = "amar",
	}

	return m_form_of.tagged_inflections({
		lang = lang, tags = {"1", "s", "pres", "ind"}, lemmas = {lemma_obj}, lemma_face = "term", POS = "verb"
	})
}

Normally, one value is returned, the formatted text, which has appended to it the formatted categories derived from the
tag-set-related categories generated by the specs in [Module:form of/cats]]. To suppress this, set
`data.no_format_categories` = {true}, in which case two values are returned, the formatted text without any formatted
categories appended and a list of the categories to be formatted.

NOTE: There are two sets of categories that may be generated: (1) categories derived directly from the tag sets, as
specified in [[Module:form of/cats]]; (2) categories derived from tag-set labels, either (a) set explicitly by the
caller in `data.tag_sets`, (b) specified by the user using `<lb:...>` attached to the last tag in a tag set, or
(c) specified in [[Module:form of/cats]]. The second type (label-related categories) are currently not returned in
the second return value of {tagged_inflections()}, and are currently inserted into the output text even if
`data.no_format_categories` is set to {true}; but they can be suppressed by setting `data.nocat` = {true} (which also
suppresses the first type of categories, those derived directly from tag sets, even if `data.no_format_categories` is
set to {true}).
]==]
function export.tagged_inflections(data)
	if not data.tags and not data.tag_sets then
		error("First argument must be a table of arguments, and `.tags` or `.tag_sets` must be specified")
	end
	if data.tags and data.tag_sets then
		error("Both `.tags` and `.tag_sets` cannot be specified")
	end
	local tag_sets = data.tag_sets
	if not tag_sets then
		tag_sets = export.split_tag_set(data.tags)
		for i, tag_set in ipairs(tag_sets) do
			tag_sets[i] = export.parse_tag_set_properties(tag_set)
		end
	end

	local inflections = {}
	local categories = {}
	for _, tag_set in ipairs(tag_sets) do
		local normalized_tag_sets = export.normalize_tag_set(tag_set.tags, data.lang, "do-track")

		for _, normalized_tag_set in ipairs(normalized_tag_sets) do
			local cur_infl = {}
			local this_categories, this_labels = export.fetch_categories_and_labels(normalized_tag_set, data.lang,
				data.POS, data.pagename, type(data.lemmas) == "table" and data.lemmas or nil)
			if not data.nocat then
				m_table.extendList(categories, this_categories)
			end
			local cur_infl = export.get_tag_set_display_form(normalized_tag_set, data.lang, data.joiner)
			if #cur_infl > 0 then
				if tag_set.labels then
					this_labels = m_table.append(tag_set.labels, this_labels)
				end
				table.insert(inflections, {infl_text = cur_infl, labels = this_labels})
			end
		end
	end

	local overall_labels, need_per_tag_set_labels
	for _, inflection in ipairs(inflections) do
		if overall_labels == nil then
			overall_labels = inflection.labels
		elseif not m_table.deepEquals(overall_labels, inflection.labels) then
			need_per_tag_set_labels = true
			overall_labels = nil
			break
		end
	end

	if not need_per_tag_set_labels then
		for _, inflection in ipairs(inflections) do
			inflection.labels = nil
		end
	end

	local format_data = m_table.shallowcopy(data)

	local function format_labels(labels, notext)
		if labels and #labels > 0 then
			return require(labels_module).show_labels { labels = labels, lang = data.lang, sort = data.sort, nocat = data.nocat } ..
				(notext and (data.pretext or "") == "" and "" or " ")
		else
			return ""
		end
	end

	local of_text = data.lemmas and " of" or ""
	local formatted_text
	if #inflections == 1 then
		if need_per_tag_set_labels then
			error("Internal error: need_per_tag_set_labels should not be set with one inflection")
		end
		format_data.text = format_labels(overall_labels, data.notext) .. (data.pretext or "") .. (data.notext and "" or
			((data.capfirst and require("Module:string utilities").ucfirst(inflections[1].infl_text) or inflections[1].infl_text) .. of_text))
		formatted_text = export.format_form_of(format_data)
	else
		format_data.text = format_labels(overall_labels, data.notext) .. (data.pretext or "") .. (data.notext and "" or
			((data.capfirst and "Inflection" or "inflection") .. of_text))
		format_data.posttext = (data.posttext or "") .. ":"
		local link = export.format_form_of(format_data)
		local text_classes = data.text_classes or "form-of-definition use-with-mention"
		for i, inflection in ipairs(inflections) do
			inflections[i] = "\n## " .. format_labels(inflection.labels, false) ..
				"<span class='" .. text_classes .. "'>" .. inflection.infl_text .. "</span>"
		end
		formatted_text = link .. table.concat(inflections)
	end

	if not data.no_format_categories then
		if #categories > 0 then
			formatted_text = formatted_text .. require("Module:utilities").format_categories(categories, data.lang,
				data.sort, nil, export.force_cat)
		end
		return formatted_text
	end
	return formatted_text, categories
end


--[==[
Given a tag set, return a flattened list all Wikidata ID's of all tags in the tag set. FIXME: Only used in a debugging
function in [[Module:se-verbs]]; move there.
]==]
function export.to_Wikidata_IDs(tag_set, lang, skip_tags_without_ids)
	local ret = {}

	local function get_wikidata_id(tag)
		local data = export.lookup_tag(tag, lang)

		if not data or not data[export.WIKIDATA] then
			if not skip_tags_without_ids then
				error('The tag "' .. tag .. '" does not have a Wikidata ID defined in the form-of data modules')
			else
				return nil
			end
		else
			return ("Q%s"):format(data[export.WIKIDATA])
		end
	end

	local normalized_tag_sets = export.normalize_tag_set(tag_set, lang)
	for _, tag_set in ipairs(normalized_tag_sets) do
		for _, tag in ipairs(tag_set) do
			if type(tag) == "table" then
				for _, subtag in ipairs(tag) do
					if type(subtag) == "table" then
						-- two-level multipart tag; FIXME: delete support for this
						for _, subsubtag in ipairs(subtag) do
							table.insert(ret, get_wikidata_id(subsubtag))
						end
					else
						table.insert(ret, get_wikidata_id(subtag))
					end
				end
			else
				table.insert(ret, get_wikidata_id(tag))
			end
		end
	end

	return ret
end


function export.dump_form_of_data(frame)
	local data = {
		data = require(export.form_of_data_module),
		data2 = require(export.form_of_data2_module)
	}
	return require("Module:JSON").toJSON(data)
end


function export.finalize_tag_data(tags, shortcuts)
	local function process_shortcut(name, shortcut)
		-- If the shortcut is already in the list, then there is a duplicate.
		if shortcuts[shortcut] then
			error("The shortcut \"" .. shortcut .. "\" (for the inflection tag \"" .. name .. "\") conflicts with an existing shortcut for the tag \"" .. shortcuts[shortcut] .. "\".")
		elseif tags[shortcut] then
			error("The shortcut \"" .. shortcut .. "\" (for the inflection tag \"" .. name .. "\") conflicts with an existing tag with that name.")
		end

		shortcuts[shortcut] = name
	end
	for name, data in pairs(tags) do
		local data_shortcuts = data[export.SHORTCUTS]
		if data_shortcuts then
			if type(data_shortcuts) == "string" then
				process_shortcut(name, data_shortcuts)
			else
				for _, shortcut in ipairs(data_shortcuts) do
					process_shortcut(name, shortcut)
				end
			end
		end
	end
end


return export