cmake-ucd

npm i cmake-ucd

find_package(cmake-ucd REQUIRED PATHS node_modules/cmake-ucd)

Fetches the Unicode Character Database at configure time and provides the helpers for turning it into the character tables of a C library.

API

CMake

ucd_version

The version of the database to fetch, as a cache variable, so that every project of a build agrees on it.

ucd_data

The directory that fetched files are written to, in the top-level build tree.

ucd_fetch(<collection> <names>... [PATHS <variable>])

Fetches the named files of a collection, appending their absolute paths to the variable named by PATHS. A file already fetched is left alone, the data of a released version never changing.

The collection names the part of unicode.org/Public to take the files from:

| Collection | Holds | | ----------- | ------------------------------------------------------------ | | UCD | The character database proper, such as UnicodeData.txt | | EXTRACTED | Properties extracted from it, such as DerivedBidiClass.txt | | IDNA | The IDNA data, such as IdnaMappingTable.txt |

The IDNA data is versioned separately and is not always given a directory of its own version: as of Unicode 17.0.0 the numbered directories stop at 16.0.0 while the 17.0.0 data is published under latest. Both are therefore tried, and a file taken from latest is checked to declare the version that was asked for rather than trusted to be it. Asking for a version that no published data holds is an error rather than a silent substitution.

JavaScript

const { open, ranges, rangeBlocks, pool, hex, table } = require('cmake-ucd')

open(dir) gives the readers that need to know where the files are:

lines(file)

Iterates the data lines, stripping comments and yielding the semicolon separated fields of each.

missing(file)

Iterates the @missing annotations, which give the values of the code points that a file leaves unlisted.

property(file, values, fallback)

Reads a file of the extracted properties into an array holding the value of every code point. values maps the names the file uses to the values to store, and a code point whose name is absent from it takes fallback, which is how the values a caller has no interest in are lumped together.

The rest take their data as arguments:

range(field) and codePoints(field)

Parse a code point or inclusive range such as 0041..005A, and a space separated sequence.

ranges(property)

Compresses an array holding the value of every code point into a sorted, gap-free list of [start, value] pairs, each extending up to the start of the pair that follows it. How the pairs are packed into a table is left to the caller, the room a value needs being its own to know.

blocks(shift), rangeBlocks(starts, shift) and entryBlocks(codePoints, shift)

Index a table by block, so that a lookup searches only the entries that may hold a given code point. shift is the number of low bits of a code point that a block leaves to the search it narrows down.

pool()

A pool of code point sequences that the entries of a table refer to by offset and length, sharing identical sequences.

hex(value) and table(entries, perLine)

Format a value, and wrap a list of already formatted entries across as many lines as needed.

License

Apache-2.0