xref: /third_party/icu/tools/cldr/cldr-to-icu/build-icu-data.xml

<!-- © 2019 and later: Unicode, Inc. and others.
     License & terms of use: http://www.unicode.org/copyright.html -->

<!--================================================================================
    Setup:
    Follow the installation instructions in README.txt in this directory.

    To build ICU data files:
    1: Determine the CLDR base directory and set the CLDR_DIR environment variable.
    2: Determine the flags required (see the list of properties below).
    3: Run: ant -f build-icu-data.xml -D<flag-name>=<flag-value>...
    ================================================================================-->
<!-- TODO: Add things like copying of a template directory and deleting previous files
     (perhaps always generate into a temporary directory and copy back to avoid having
      inconsistent state when the conversion is cancelled). -->
<project name="Convert" default="all" basedir="." xmlns:if="ant:if" xmlns:unless="ant:unless">

    <target name="all" depends="init-args, prepare-jar, clean, convert"/>

    <!-- Initialize the properties which were not already set on the command line. -->
    <target name="init-args">
        <property environment="env"/>
        <!-- Inherit properties from environment variable unless specified. As usual
             with Ant, this is messier than it should be. All we are saying here is:
             "Use the property if explicitly set, otherwise use the environment variable."
             We cannot just set the property to the environment variable, since expansion
             fails for non existent properties, and you are left with a literal value of
             "${env.CLDR_DATA_DIR}". -->
        <condition property="cldrDataDir" value="${env.CLDR_DATA_DIR}">
            <isset property="env.CLDR_DATA_DIR"/>
        </condition>
        <fail unless="cldrDataDir"
              message="Set the CLDR_DATA_DIR environment variable (or cldrDataDir property) to the CLDR data directory (typically ending in '/production')"/>

        <!-- Ant does not inherit this from the user's environment (and it can matter).
             This is only needed because we have to "exec" a new Ant task below. -->
        <condition property="javaHome" value="${env.JAVA_HOME}">
            <isset property="env.JAVA_HOME"/>
        </condition>

        <!-- The output directory into which to write the converted ICU data. By default
             this will overwrite (without deletion) the ICU data files in this ICU release,
             so it is recommended that for testing, it be set to another value.  -->
        <property name="outDir" value="${basedir}/../../../icu4c/source/data/"/>

        <!-- The output directory into which to write generated C/C++ code.  By default
             this will overwrite (without deletion) the generated C/C++ files in this
             ICU release, so it is recommended that for testing, it be set to another value. -->
        <property name="genCCodeDir" value="${basedir}/../../../icu4c/source/"/>

        <!-- The output directory into which to write generated Java code.  By default
             this will overwrite (without deletion) the generated Java files in this
             ICU release, so it is recommended that for testing, it be set to another value. -->
        <property name="genJavaCodeDir" value="${basedir}/../../../icu4j/main/classes/core"/>
        
        <!-- Set this to true to prevent build-icu-data.xml from generating the generated
             ICU source files -->
        <property name="dontGenCode" value="false" />

        <!-- The directory in which the additional ICU XML data is stored. -->
        <property name="specialsDir" value="${basedir}/../../../icu4c/source/data/xml"/>

        <!-- Default value for ICU version (icuver.txt). Update this for each release. -->
        <property name="icuVersion" value="72.1.0.0"/>

        <!-- Default value for ICU data version (icuver.txt). Update this for each release. -->
        <property name="icuDataVersion" value="72.1.0.0"/>

        <!-- An override for the CLDR version string (icuver.txt and others). This will be
             extracted from the CLDR library used for building the data if not set here. -->
        <property name="cldrVersion" value=""/>

        <!-- The minimum draft status for CLDR data to be used in the conversion. See
             CldrDraftStatus for more details. -->
        <property name="minDraftStatus" value="contributed"/>

        <!-- A regular expression to match the locale IDs to be generated (useful for
             debugging specific regions). This is applied after locale ID specifications
             have been expanded into full locale IDs, so the value "en" will NOT match
             "en_GB" or "en_001" etc. -->
        <property name="localeIdFilter" value=""/>

        <!-- Whether to synthetically generate "pseudo locale" data ("en_XA" and "ar_XB"). -->
        <property name="includePseudoLocales" value="false"/>

        <!-- Whether to emit a debug report containing some possibly useful information after
             the conversion has finished. -->
        <!-- TODO: Currently this isn't hugely useful, so find out what people want. -->
        <property name="emitReport" value="false"/>

        <!-- List of output "types" to be generated (e.g. "rbnf,plurals,locales"); an empty
             list means "build everything".

             Note that the grouping of types is based on the legacy converter behaviour and
             is not always directly associated with an output directory (e.g. "locales"
             produces locale data for curr/, lang/, main/, region/, unit/, zone/ but NOT
             coll/, brkitr/ or rbnf/).

             Pass in the value "HELP" (or any invalid value) to see the full list of types. -->
        <!-- TODO: Find out what common use cases are and use them. -->
        <property name="outputTypes" value=""/>

        <!-- Override to force the 'clean' task to delete files it cannot determine to be
             auto-generated by this tool. This is useful if the file header changes since
             the heading is what's used to recognize auto-generated files. -->
        <property name="forceDelete" value="false"/>
    </target>

    <!-- Build a standalone JAR which is called by Ant (and which avoids needing to mess
         about making Ant know the Maven class-path). -->
    <target name="prepare-jar" depends="init-args">
        <exec executable="mvn" searchpath="true" failonerror="true">
            <arg value="compile"/>
        </exec>
    </target>

    <!-- Somewhat hacky wrapper target which invokes the real conversion task.
         This is done so we can set the environment variable of the new process and
         effectively overwrite the CLDR_DIR value. If ever the CLDR library doesn't
         need to use CLDR_DIR at runtime to find the production data, this can all be
         removed. -->
    <target name="convert" depends="init-args, prepare-jar">
        <exec executable="ant" searchpath="true" failonerror="true">
            <!-- The CLDR library wants CLDR_DIR set, to the data directory. -->
            <env key="CLDR_DIR" value="${cldrDataDir}" />
            <!-- Force inherit JAVA_HOME (this can be important). -->
            <env key="JAVA_HOME" value="${javaHome}" />
            <!-- Initial Ant command line with all the "interesting" bit in. -->
            <arg line="-f build-icu-data.xml convert-impl -DcldrDir=${cldrDataDir}"/>
            <!-- List all properties in the "convert-impl" task (except cldrDir). -->
            <arg value="-DoutDir=${outDir}"/>
            <arg value="-DgenCCodeDir=${genCCodeDir}"/>
            <arg value="-DgenJavaCodeDir=${genJavaCodeDir}"/>
            <arg value="-DdontGenCode=${dontGenCode}"/>
            <arg value="-DspecialsDir=${specialsDir}"/>
            <arg value="-DoutputTypes=${outputTypes}"/>
            <arg value="-DicuVersion=${icuVersion}"/>
            <arg value="-DicuDataVersion=${icuDataVersion}"/>
            <arg value="-DcldrVersion=${cldrVersion}"/>
            <arg value="-DminDraftStatus=${minDraftStatus}"/>
            <arg value="-DlocaleIdFilter=${localeIdFilter}"/>
            <arg value="-DincludePseudoLocales=${includePseudoLocales}"/>
            <arg value="-DemitReport=${emitReport}"/>
        </exec>
    </target>

    <!-- Do the actual CLDR data conversion, based on the command line arguments, built in
         default properties and the configuration in the "<convert>" element below. -->
    <target name="convert-impl">
        <taskdef name="convert" classname="org.unicode.icu.tool.cldrtoicu.ant.ConvertIcuDataTask">
            <classpath>
                <pathelement path="target/cldr-to-icu-1.0-SNAPSHOT-jar-with-dependencies.jar"/>
            </classpath>
        </taskdef>
        <taskdef name="generateCode" classname="org.unicode.icu.tool.cldrtoicu.ant.GenerateCodeTask">
            <classpath>
                <pathelement path="target/cldr-to-icu-1.0-SNAPSHOT-jar-with-dependencies.jar"/>
            </classpath>
        </taskdef>
        <convert cldrDir="${cldrDir}" outputDir="${outDir}" specialsDir="${specialsDir}"
                 outputTypes="${outputTypes}" cldrVersion="${cldrVersion}"
                 icuVersion="${icuVersion}" icuDataVersion="${icuDataVersion}"
                 minimalDraftStatus="${minDraftStatus}" localeIdFilter="${localeIdFilter}"
                 includePseudoLocales="${includePseudoLocales}" emitReport="${emitReport}">

            <!-- The primary set of locale IDs to be generated by default. The IDs in this list are
                 automatically expanded to include default scripts and all available regions. The
                 rules are:

                 1) Base languages are expanded to include default scripts (e.g. "en" -> "en_Latn").
                 2) All region and variant subtags are added for any base language or language+script
                    (e.g. "en" -> "en_GB" or "shi_Latn" -> "shi_Latn_MA").

                 If a non-default script is desired it should be listed explicitly (e.g. "sr_Latn").

                 Locale IDs with deprecated subtags (which become aliases) must still be listed in
                 full (e.g. "en_RH" or "sr_Latn_YU").
            -->
            <localeIds>
                // A
                af, agq, ak, am, ar, ars, as, asa, ast, az, az_AZ, az_Cyrl

                // B
                bas, be, bem, bez, bg, bgc, bho, bm, bn, bo, br, brx, bs, bs_BA, bs_Cyrl

                // C
                ca, ccp, ce, ceb, cgg, chr, ckb, cs, cv, cy

                // D
                da, dav, de, dje, doi, dsb, dua, dyo, dz

                // E
                ebu, ee, el, en, en_NH, en_RH, eo, es, et, eu, ewo

                // F
                fa, ff, ff_Adlm, ff_CM, ff_GN, ff_MR, ff_SN, fi, fil, fo, fr, fur, fy

                // G
                ga, gd, gl, gsw, gu, guz, gv

                // H
                ha, haw, he, hi, hi_Latn, hr, hsb, hu, hy

                // I
                ia, id, ig, ii, in, in_ID, is, it, iw, iw_IL

                // J
                ja, jgo, jmc, jv

                // K
                ka, kab, kam, kde, kea, kgp, khq, ki, kk, kkj, kl, kln, km, kn, ko, kok, ks
                ks_Deva, ks_IN, ksb, ksf, ksh, ku, kw, ky

                // L
                lag, lb, lg, lkt, ln, lo, lrc, lt, lu, luo, luy, lv

                // M
                mai, mas, mer, mfe, mg, mgh, mgo, mi, mk, ml, mn, mni, mni_IN, mo, mr, ms
                mt, mua, my, mzn

                // N
                naq, nb, nd, ne, nl, nmg, nn, nnh, no, no_NO, no_NO_NY, nus, nyn

                // O
                om, or, os

                // P
                pa, pa_Arab, pa_IN, pa_PK, pcm, pl, ps, pt

                // Q
                qu

                // R
                raj, rm, rn, ro, rof, ru, rw, rwk

                // S
                sa, sah, saq, sat, sat_IN, sbp, sc, sd, sd_Deva, sd_IN, sd_PK, se, seh, ses, sg, sh, sh_BA, sh_CS, sh_YU
                shi, shi_Latn, shi_MA, si, sk, sl, smn, sn, so, sq, sr, sr_BA, sr_CS, sr_Cyrl_CS, sr_Cyrl_YU, sr_Latn
                sr_Latn_CS, sr_Latn_YU, sr_ME, sr_RS, sr_XK, sr_YU, su, su_ID, sv, sw

                // T
                ta, te, teo, tg, th, ti, tk, tl, tl_PH, to, tr, tt, twq, tzm

                // U
                ug, uk, ur, uz, uz_AF, uz_Arab, uz_Cyrl, uz_UZ

                // V
                vai, vai_LR, vai_Latn, vi, vun

                // W
                wae, wo

                // X
                xh, xog

                // Y
                yav, yi, yo, yrl, yue, yue_CN, yue_HK, yue_Hans

                // Z
                zgh, zh, zh_CN, zh_HK, zh_Hant, zh_MO, zh_SG, zh_TW, zu
            </localeIds>

            <!-- The following elements configure directories in which a subset of the available
                 locales IDs should be generated. Unlike the main <localeId> element, these
                 filters must specify all locale IDs in full (but since they mostly select base
                 languages, this isn't a big deal).

                 As well as allowing some data directories to have a subset of available data (via
                 the <localeIds> element) there are also mechanisms for controlling aliasing and
                 the locale parent relation which allows the sharing of some ICU data in cases
                 where it would otherwise need to be copied. The two mechanisms are:

                 1: inheritLanguageSubtag: Used to rewrite the parent of a locale ID from "root" to
                    its language subtag (e.g. "zh_Hant" has a natural parent of "root", but to allow
                    some base language data to be shared it can be made to have a parent of "zh").

                 2: forcedAlias: Used to add aliases for specific directories in order to affect the
                    ICU behaviour in special cases.

                 Between them these mechanisms are known as "tailorings" of the affected locales. -->
            <!-- TODO: Explain why these special cases are needed/different. -->

            <!-- Collation data is large, but also more sharable than other data, which is why there
                 are a number of aliases and parent remappings for this directory. -->
            <directory dir="coll" inheritLanguageSubtag="bs_Cyrl, sr_Latn, zh_Hant">
                <!-- These aliases are to avoid needing to copy and maintain the same collation data
                     for "zh" and "yue". The maximized versions of "yue_Hans" is "yue_Hans_CN" (vs
                     "zh_Hans_CN"), and for "yue" it's "yue_Hant_HK" (vs "zh_Hant_HK"), so the
                     aliases are effectively just rewriting the base language. -->
                <forcedAlias source="yue" target="zh_Hant"/>
                <forcedAlias source="yue_Hant" target="zh_Hant"/>
                <forcedAlias source="yue_CN" target="zh_Hans"/>
                <forcedAlias source="yue_Hans" target="zh_Hans"/>
                <forcedAlias source="yue_Hans_CN" target="zh_Hans"/>

                <!-- TODO: Find out and document this properly. -->
                <forcedAlias source="sr_ME" target="sr_Cyrl_ME"/>

                <localeIds>
                    root,

                    // A-B
                    af, am, ars, ar, as, az, be, bg, bn, bo, br, bs_Cyrl, bs,

                    // C-F
                    ca, ceb, chr, cs, cy, da, de_AT, de, dsb, dz, ee, el, en,
                    en_US_POSIX, en_US, eo, es, et, fa_AF, fa, ff_Adlm, ff, fil, fi, fo, fr_CA, fr, fy,

                    // G-J
                    ga, gl, gu, ha, haw, he, hi, hr, hsb, hu, hy,
                    id_ID, id, ig, in, in_ID, is, it, iw_IL, iw, ja,

                    // K-P
                    ka, kk, kl, km, kn, kok, ko, ku, ky, lb, lkt, ln, lo, lt, lv,
                    mk, ml, mn, mo, mr, ms, mt, my, nb, nb_NO, ne, nl, nn, no, no_NO,
                    om, or, pa_IN, pa, pa_Guru, pl, ps, pt,

                    // R-T
                    ro, ru, sa, se, sh_BA, sh_CS, sh, sh_YU, si, sk, sl, smn, sq,
                    sr_BA, sr_Cyrl_ME, sr_Latn, sr_ME, sr_RS, sr, sv, sw,
                    ta, te, th, tk, to, tr,

                    // U-Z
                    ug, uk, ur, uz, vi, wae, wo, xh, yi, yo, yue_CN, yue_Hans_CN, yue_Hans
                    yue_Hant, yue, zh_CN, zh_Hans, zh_Hant, zh_HK, zh_MO, zh_SG, zh_TW, zh, zu
                </localeIds>
            </directory>

            <directory dir="rbnf">
                <!-- It is not at all clear why this is being done. It's certainly not exactly the
                     same as above, since (a) the alias is reversed (b) "zh_Hant" does exist, with
                     different data than "yue", so this alias is not just rewriting the base
                     language. -->
                <!-- TODO: Find out and document this properly. -->
                <forcedAlias source="zh_Hant_HK" target="yue"/>

                <localeIds>
                    root,

                    // A-E
                    af, ak, am, ars, ar, az, be, bg, bs, ca, ccp, chr, cs, cy,
                    da, de_CH, de, ee, el, en_001, en_IN, en, eo, es_419, es_DO,
                    es_GT, es_HN, es_MX, es_NI, es_PA, es_PR, es_SV, es, es_US, et,

                    // F-P
                    fa_AF, fa, ff, fil, fi, fo, fr_BE, fr_CH, fr, ga, he, hi, hr,
                    hu, hy, id, in, is, it, iw, ja, ka, kk, kl, km, ko, ky, lb,
                    lo, lrc, lt, lv, mk, ms, mt, my, nb, ne, nl, nn, no, pl, pt_PT, pt,

                    // Q-Z
                    qu, ro, ru, se, sh, sk, sl, sq, sr_Latn, sr, su, sv, sw, ta, th, tr,
                    uk, vi, yue_Hans, yue, zh_Hant_HK, zh_Hant, zh_HK, zh_MO, zh_TW, zh
                </localeIds>
            </directory>

            <directory dir="brkitr" inheritLanguageSubtag="zh_Hant">
                <localeIds>
                    root,
                    de, el, en, en_US_POSIX, en_US, es, fi, fr, it, ja, pt, ru, sv, zh_Hant, zh
                </localeIds>
            </directory>

            <!-- GLOBAL ALIASES -->

            <!-- Some spoken languages (e.g. "ars") inherit all their data from a written language
                 (e.g. "ar_SA"). However CLDR doesn't currently support a way to represent that
                 relationship. Unlike deprecated languages for which an alias can be inferred from
                 the "languageAlias" CLDR data, there's no way in CLDR to represent the fact that
                 we want "ars" (a non-deprecated language) to inherit the data of "ar_SA".

                 This alias is the first example of potentially many cases where ICU needs to
                 generate an alias in order to affect "sideways inheritance" for spoken languages,
                 and at some stage it should probably be supported properly in the CLDR data. -->
            <forcedAlias source="ars" target="ar_SA"/>

            <!-- A legacy global alias (note that "no_NO_NY" is not even structurally valid). -->
            <forcedAlias source="no_NO_NY" target="nn_NO"/>

            <!-- This one is a bit silly, it is just to generate a stub for no_NO, which is
                 not in CLDR. If we do not do this, then including it in localeIds will generate
                 empty no_Latn and no_Latn_NO and then no_NO aliasing to no_Latn_NO. -->
            <forcedAlias source="no_NO" target="no"/>

            <!-- ALTERNATE VALUES -->

            <!-- The following elements configure alternate values for some special case paths.
                 The target path will only be replaced if both it, and the source path, exist in
                 the CLDR data (paths will not be modified if only the source path exists).

                 Since the paths must represent the same semantic type of data, they must be in the
                 same "namespace" (same element names) and must not contain value attributes. Thus
                 they can only differ by distinguishing attributes (either added or modified).

                 This feature is typically used to select alternate translations (e.g. short forms)
                 for certain paths. -->
            <!-- <altPath target="//path/to/value[@attr='foo']"
                          source="//path/to/value[@attr='bar']"
                          locales="xx,yy_ZZ"/> -->
        </convert>
        
        <generateCode cldrDir="${cldrDir}" cOutDir="${genCCodeDir}" javaOutDir="${genJavaCodeDir}" unless:true="${dontGenCode}" />
    </target>

    <target name="clean" depends="init-args, prepare-jar">
        <taskdef name="outputDirectories" classname="org.unicode.icu.tool.cldrtoicu.ant.CleanOutputDirectoryTask">
            <classpath>
                <pathelement path="target/cldr-to-icu-1.0-SNAPSHOT-jar-with-dependencies.jar"/>
            </classpath>
        </taskdef>
        <taskdef name="generateCode" classname="org.unicode.icu.tool.cldrtoicu.ant.GenerateCodeTask">
            <classpath>
                <pathelement path="target/cldr-to-icu-1.0-SNAPSHOT-jar-with-dependencies.jar"/>
            </classpath>
        </taskdef>

        <!-- If a directory is listed here, then every file in it is assumed to be automatically
             generated by the conversion tool, unless it is explicitly listed in a <retain> element.
             The tool then checks every file to determine if it has the expected header present,
             indiciating that it was automatically generated, before deleting it.
             
             If unexpected files are found, the "clean" task will fail without deleting anything
             (unless'forceDelete' is set to override this). Note that even if 'forceDelete' is set,
             the files listed explicitly below will never be deleted by this process.

             This two-step approach minimizes the risk that the conversion process will ever
             accidentally delete a manually maintained file.             
             -->
        <outputDirectories root="${outDir}" forceDelete="${forceDelete}">
            <dir name="brkitr">
                <retain path="dictionaries"/>
                <retain path="lstm"/>
                <retain path="rules"/>
            </dir>
            <dir name="coll">
                <!-- Legacy files whose file names aren't supported for automatic generation.
                     Simple to maintain manually and unlikely to ever change again. -->
                <retain path="de__PHONEBOOK.txt"/>
                <retain path="de_.txt"/>
                <retain path="es__TRADITIONAL.txt"/>
                <retain path="es_.txt"/>
            </dir>
            <dir name="curr"/>
            <dir name="lang"/>
            <dir name="locales"/>
            <dir name="misc">
                <!-- Machine generated files produced by different tools.
                     Possibly worth moving into the new LDML conversion tool one day. -->
                <retain path="currencyNumericCodes.txt"/>
                <retain path="zoneinfo64.txt"/>
                <!-- Project file (not ICU data), unlikely to ever be auto-generated. -->
                <retain path="icudata.rc"/>
                <!-- Small high-level metadata file, stable and easy to maintain manually. -->
                <retain path="icustd.txt"/>
            </dir>
            <dir name="rbnf"/>
            <dir name="region"/>
            <dir name="translit">
                <!-- Small, easy to maintain, special case top-level files. -->
                <retain path="en.txt"/>
                <retain path="el.txt"/>
            </dir>
            <dir name="unit"/>
            <dir name="zone">
                <!-- Manually edited to support TZ database name compatibility. -->
                <retain path="tzdbNames.txt"/>
            </dir>
        </outputDirectories>
        
        <generateCode cOutDir="${genCCodeDir}" javaOutDir="${genJavaCodeDir}" action="clean" />
    </target>
</project>