blob: 3a703cb10b6a091cb39eb33b96a4ce7b18d0ae53 [file] [log] [blame]
# Copyright (c) 2012-2015 International Business Machines
# Corporation and others. All Rights Reserved.
#
# This file should be in UTF-8 with a signature byte sequence ("BOM").
#
# collationtest.txt: Collation test data.
#
# created on: 2012apr13
# created by: Markus W. Scherer
# A line with "** test: description" is used for verbose and error output.
# A collator can be set with "@ root" or "@ locale language-tag",
# for example "@ locale de-u-co-phonebk".
# An old-style locale ID can also be used, for example "@ locale de@collation=phonebook".
# A collator can be built with "@ rules".
# An "@ rules" line is followed by one or more lines with the tailoring rules.
# A collator can be modified with "% attribute=value".
# "* compare" tests the order (= or <) of the following strings.
# The relation can be "=" or "<" (the level of the difference is not specified)
# or "<1", "<2", "<c", "<3", "<4" (indicating the level of the difference).
# Test sections ("* compare") are terminated by
# definitions of new collators, changing attributes, or new test sections.
** test: simple CEs & expansions
# Many types of mappings are tested elsewhere, including via the UCA conformance tests.
# Here we mostly cover a few unusual mappings.
@ rules
&\x01 # most control codes are ignorable
<<<\u0300 # tertiary CE
&9<\x00 # NUL not ignorable
&\uA00A\uA00B=\uA002 # two long-primary CEs
&\uA00A\uA00B\u00050005=\uA003 # three CEs, require 64 bits
* compare
= \x01
= \x02
<3 \u0300
<1 9
<1 \x00
= \x01\x00\x02
<1 a
<3 a\u0300
<2 a\u0308
= ä
<1 b
<1 か # Hiragana Ka (U+304B)
<2 か\u3099 # plus voiced sound mark
= が # Hiragana Ga (U+304C)
<1 \uA00A\uA00B
= \uA002
<1 \uA00A\uA00B\u00050004
<1 \uA00A\uA00B\u00050005
= \uA003
<1 \uA00A\uA00B\u00050006
** test: contractions
# Create some interesting mappings, and map some normalization-inert characters
# (which are not subject to canonical reordering)
# to some of the same CEs to check the sequence of CEs.
@ rules
# Contractions starting with 'a' should not continue with any character < U+0300
# so that we can test a shortcut for that.
&a=ⓐ
&b<bz=ⓑ
&d<dz\u0301=ⓓ # d+z+acute
&z
<a\u0301=Ⓐ # a+acute sorts after z
<a\u0301\u0301=Ⓑ # a+acute+acute
<a\u0301\u0301\u0358=Ⓒ # a+acute+acute+dot above right
<a\u030a=Ⓓ # a+ring
<a\u0323=Ⓔ # a+dot below
<a\u0323\u0358=Ⓕ # a+dot below+dot above right
<a\u0327\u0323\u030a=Ⓖ # a+cedilla+dot below+ring
<a\u0327\u0323bz=Ⓗ # a+cedilla+dot below+b+z
&\U0001D158=⁰ # musical notehead black (has a symbol primary)
<\U0001D158\U0001D165=¼ # musical quarter note
# deliberately missing prefix contractions:
# dz
# a\u0327
# a\u0327\u0323
# a\u0327\u0323b
&\x01
<<<\U0001D165=¹ # musical stem (ccc=216)
<<<\U0001D16D=² # musical augmentation dot (ccc=226)
<<<\U0001D165\U0001D16D=³ # stem+dot (ccc=216 226)
&\u0301=❶ # acute (ccc=230)
&\u030a=❷ # ring (ccc=230)
&\u0308=❸ # diaeresis (ccc=230)
<<\u0308\u0301=❹ # diaeresis+acute (=dialytika tonos) (ccc=230 230)
&\u0327=❺ # cedilla (ccc=202)
&\u0323=❻ # dot below (ccc=220)
&\u0331=❼ # macron below (ccc=220)
<<\u0331\u0358=❽ # macron below+dot above right (ccc=220 232)
&\u0334=❾ # tilde overlay (ccc=1)
&\u0358=❿ # dot above right (ccc=232)
&\u0f71=① # tibetan vowel sign aa
&\u0f72=② # tibetan vowel sign i
# \u0f71\u0f72 # tibetan vowel sign aa + i = ii = U+0F73
&\u0f73=③ # tibetan vowel sign ii (ccc=0 but lccc=129)
** test: simple contractions
# Some strings are chosen to cause incremental contiguous contraction matching to
# go into partial matches for prefixes of contractions
# (where the prefixes are deliberately not also contractions).
# When there is no complete match, then the matching code must back out of those
# so that discontiguous contractions work as specified.
* compare
# contraction starter with no following text, or mismatch, or blocked
<1 a
= ⓐ
<1 aa
= ⓐⓐ
<1 ab
= ⓐb
<1 az
= ⓐz
* compare
<1 a
<2 a\u0308\u030a # ring blocked by diaeresis
= ⓐ❸❷
<2 a\u0327
= ⓐ❺
* compare
<2 \u0308
= ❸
<2 \u0308\u030a\u0301 # acute blocked by ring
= ❸❷❶
* compare
<1 \U0001D158
= ⁰
<1 \U0001D158\U0001D165
= ¼
# no discontiguous contraction because of missing prefix contraction d+z,
# and a starter ('z') after the 'd'
* compare
<1 dz\u0323\u0301
= dz❻❶
# contiguous contractions
* compare
<1 abz
= ⓐⓑ
<1 abzz
= ⓐⓑz
* compare
<1 a
<1 z
<1 a\u0301
= Ⓐ
<1 a\u0301\u0301
= Ⓑ
<1 a\u0301\u0301\u0358
= Ⓒ
<1 a\u030a
= Ⓓ
<1 a\u0323\u0358
= Ⓕ
<1 a\u0327\u0323\u030a # match despite missing prefix
= Ⓖ
<1 a\u0327\u0323bz
= Ⓗ
* compare
<2 \u0308\u0308\u0301 # acute blocked from first diaeresis, contracts with second
= ❸❹
* compare
<1 \U0001D158\U0001D165
= ¼
* compare
<3 \U0001D165\U0001D16D
= ³
** test: discontiguous contractions
* compare
<1 a\u0327\u030a # a+ring skips cedilla
= Ⓓ❺
<2 a\u0327\u0327\u030a # a+ring skips 2 cedillas
= Ⓓ❺❺
<2 a\u0327\u0327\u0327\u030a # a+ring skips 3 cedillas
= Ⓓ❺❺❺
<2 a\u0334\u0327\u0327\u030a # a+ring skips tilde overlay & 2 cedillas
= Ⓓ❾❺❺
<1 a\u0327\u0323 # a+dot below skips cedilla
= Ⓔ❺
<1 a\u0323\u0301\u0358 # a+dot below+dot ab.r.: 2-char match, then skips acute
= Ⓕ❶
<2 a\u0334\u0323\u0358 # a+dot below skips tilde overlay
= Ⓕ❾
* compare
<2 \u0331\u0331\u0358 # macron below+dot ab.r. skips the second macron below
= ❽❼
* compare
<1 a\u0327\u0331\u0323\u030a # a+ring skips cedilla, macron below, dot below (dot blocked by macron)
= Ⓓ❺❼❻
<1 a\u0327\u0323\U0001D16D\u030a # a+dot below skips cedilla
= Ⓔ❺²❷
<2 a\u0327\u0327\u0323\u030a # a+dot below skips 2 cedillas
= Ⓔ❺❺❷
<2 a\u0327\u0323\u0323\u030a # a+dot below skips cedilla
= Ⓔ❺❻❷
<2 a\u0334\u0327\u0323\u030a # a+dot below skips tilde overlay & cedilla
= Ⓔ❾❺❷
* compare
<1 \U0001D158\u0327\U0001D165 # quarter note skips cedilla
= ¼❺
<1 a\U0001D165\u0323 # a+dot below skips stem
= Ⓔ¹
# partial contiguous match, backs up, matches discontiguous contraction
<1 a\u0327\u0323b
= Ⓔ❺b
<1 a\u0327\u0323ba
= Ⓔ❺bⓐ
# a+acute+acute+dot above right skips cedilla, continues matching 2 same-ccc combining marks
* compare
<1 a\u0327\u0301\u0301\u0358
= Ⓒ❺
# FCD but not NFD
* compare
<1 a\u0f73\u0301 # a+acute skips tibetan ii
= Ⓐ③
# FCD but the 0f71 inside the 0f73 must be skipped
# to match the discontiguous contraction of the first 0f71 with the trailing 0f72 inside the 0f73
* compare
<1 \u0f71\u0f73 # == \u0f73\u0f71 == \u0f71\u0f71\u0f72
= ③①
** test: discontiguous contractions with nested contractions
* compare
<1 a\u0323\u0308\u0301\u0358
= Ⓕ❹
<2 a\u0323\u0308\u0301\u0308\u0301\u0358
= Ⓕ❹❹
** test: discontiguous contractions with interleaved contractions
* compare
# a+ring & cedilla & macron below+dot above right
<1 a\u0327\u0331\u030a\u0358
= Ⓓ❺❽
# a+ring & 1x..3x macron below+dot above right
<2 a\u0331\u030a\u0358
= Ⓓ❽
<2 a\u0331\u0331\u030a\u0358\u0358
= Ⓓ❽❽
# also skips acute
<2 a\u0331\u0331\u0331\u030a\u0301\u0358\u0358\u0358
= Ⓓ❽❽❽❶
# a+dot below & stem+augmentation dot, followed by contiguous d+z+acute
<1 a\U0001D165\u0323\U0001D16Ddz\u0301
= Ⓔ³ⓓ
** test: some simple string comparisons
@ root
* compare
# first string compares against ""
= \u0000
< a
<1 b
<3 B
= \u0000B\u0000
** test: compare with strength=primary
% strength=primary
* compare
<1 a
<1 b
= B
** test: compare with strength=secondary
% strength=secondary
* compare
<1 a
<1 b
= B
** test: compare with strength=tertiary
% strength=tertiary
* compare
<1 a
<1 b
<3 B
** test: compare with strength=quaternary
% strength=quaternary
* compare
<1 a
<1 b
<3 B
** test: compare with strength=identical
% strength=identical
* compare
<1 a
<1 b
<3 B
** test: côté with forwards secondary
@ root
* compare
<1 cote
<2 coté
<2 côte
<2 côté
** test: côté with forwards secondary vs. U+FFFE merge separator
# Merged sort keys: On each level, any difference in the first segment
# must trump any further difference.
* compare
<1 cote\uFFFEcôté
<2 coté\uFFFEcôte
<2 côte\uFFFEcoté
<2 côté\uFFFEcote
** test: côté with backwards secondary
% backwards=on
* compare
<1 cote
<2 côte
<2 coté
<2 côté
** test: côté with backwards secondary vs. U+FFFE merge separator
# Merged sort keys: On each level, any difference in the first segment
# must trump any further difference.
* compare
<1 cote\uFFFEcôté
<2 côte\uFFFEcoté
<2 coté\uFFFEcôte
<2 côté\uFFFEcote
** test: U+FFFE on identical level
@ root
% strength=identical
* compare
# All of these control codes are completely-ignorable, so that
# their low code points are compared with the merge separator.
# The merge separator must compare less than any other character.
<1 \uFFFE\u0001\u0002\u0003
<i \u0001\uFFFE\u0002\u0003
<i \u0001\u0002\uFFFE\u0003
<i \u0001\u0002\u0003\uFFFE
* compare
# The merge separator must even compare less than U+0000.
<1 \uFFFE\u0000\u0000
<i \u0000\uFFFE\u0000
<i \u0000\u0000\uFFFE
** test: Hani < surrogates < U+FFFD
# Note: compareUTF8() treats unpaired surrogates like U+FFFD,
# so with that the strings with surrogates will compare equal to each other
# and equal to the string with U+FFFD.
@ root
% strength=identical
* compare
<1 abz
<1 a\u4e00z
<1 a\U00020000z
<1 a\ud800z
<1 a\udbffz
<1 a\udc00z
<1 a\udfffz
<1 a\ufffdz
** test: script reordering
@ root
% reorder Hani Zzzz digit
* compare
<1 ?
<1 +
<1 丂
<1 a
<1 α
<1 5
% reorder default
* compare
<1 ?
<1 +
<1 5
<1 a
<1 α
<1 丂
** test: empty rules
@ rules
* compare
<1 a
<2 ä
<3 Ä
<1 b
** test: very simple rules
@ rules
&a=e<<<<q<<<<r<x<<<X<<y<<<Y;z,Z
% strength=quaternary
* compare
<1 a
= e
<4 q
<4 r
<1 x
<3 X
<2 y
<3 Y
<2 z
<3 Z
** test: tailoring twice before a root position: primary
@ rules
&[before 1]b<p
&[before 1]b<q
* compare
<1 a
<1 p
<1 q
<1 b
** test: tailoring twice before a root position: secondary
@ rules
&[before 2]ſ<<p
&[before 2]ſ<<q
* compare
<1 s
<2 p
<2 q
<2 ſ
# secondary-before common weight
@ rules
&[before 2]b<<p
&[before 2]b<<q
* compare
<1 a
<1 p
<2 q
<2 b
** test: tailoring twice before a root position: tertiary
@ rules
&[before 3]B<<<p
&[before 3]B<<<q
* compare
<1 b
<3 p
<3 q
<3 B
# tertiary-before common weight
@ rules
&[before 3]b<<<p
&[before 3]b<<<q
* compare
<1 a
<1 p
<3 q
<3 b
@ rules
&[before 2]b<<s
&[before 3]s<<<p
&[before 3]s<<<q
* compare
<1 a
<1 p
<3 q
<3 s
<2 b
** test: tailor after completely ignorable
@ rules
&\x00<<<x<<y
* compare
= \x00
= \x1F
<3 x
<2 y
** test: secondary tailoring gaps, ICU ticket 9362
@ rules
&[before 2]s<<'_'
&s<<r # secondary between s and ſ (long s)
&ſ<<*a-q # more than 15 between ſ and secondary CE boundary
&[before 2][first primary ignorable]<<u<<v # between secondary CE boundary & lowest secondary CE
&[last primary ignorable]<<y<<z
* compare
<2 u
<2 v
<2 \u0332 # lowest secondary CE
<2 \u0308
<2 y
<2 z
<1 s_
<2 ss
<2 sr
<2 sſ
<2 sa
<2 sb
<2 sp
<2 sq
<2 sus
<2 svs
<2 rs
** test: tertiary tailoring gaps, ICU ticket 9362
@ rules
&[before 3]t<<<'_'
&t<<<r # tertiary between t and fullwidth t
&ᵀ<<<*a-q # more than 15 between ᵀ (modifier letter T) and tertiary CE boundary
&[before 3][first secondary ignorable]<<<u<<<v # between tertiary CE boundary & lowest tertiary CE
&[last secondary ignorable]<<<y<<<z
* compare
<3 u
<3 v
# Note: The root collator currently does not map any characters to tertiary CEs.
<3 y
<3 z
<1 t_
<3 tt
<3 tr
<3 tt
<3 tᵀ
<3 ta
<3 tb
<3 tp
<3 tq
<3 tut
<3 tvt
<3 rt
** test: secondary & tertiary around root character
@ rules
&[before 2]m<<r
&m<<s
&[before 3]m<<<u
&m<<<v
* compare
<1 l
<1 r
<2 u
<3 m
<3 v
<2 s
<1 n
** test: secondary & tertiary around tailored item
@ rules
&m<x
&[before 2]x<<r
&x<<s
&[before 3]x<<<u
&x<<<v
* compare
<1 m
<1 r
<2 u
<3 x
<3 v
<2 s
<1 n
** test: more nesting of secondary & tertiary before
@ rules
&[before 3]m<<<u
&[before 2]m<<r
&[before 3]r<<<q
&m<<<w
&m<<t
&[before 3]w<<<v
&w<<<x
&w<<s
* compare
<1 l
<1 q
<3 r
<2 u
<3 m
<3 v
<3 w
<3 x
<2 s
<2 t
<1 n
** test: case bits
@ rules
&w<x # tailored CE getting case bits
=uv=uV=Uv=UV # 2 chars -> 1 CE
&ae=ch=cH=Ch=CH # 2 chars -> 2 CEs
&rst=yz=yZ=Yz=YZ # 2 chars -> 3 CEs
% caseFirst=lower
* compare
<1 ae
= ch
<3 cH
<3 Ch
<3 CH
<1 rst
= yz
<3 yZ
<3 Yz
<3 YZ
<1 w
<1 x
= uv
<3 uV
= Uv # mixed case on single CE cannot distinguish variations
<3 UV
** test: tertiary CEs, tertiary, caseLevel=off, caseFirst=lower
@ rules
&\u0001<<<t<<<T # tertiary CEs
% caseFirst=lower
* compare
<1 aa
<3 aat
<3 aaT
<3 aA
<3 aAt
<3 ata
<3 aTa
** test: tertiary CEs, tertiary, caseLevel=off, caseFirst=upper
% caseFirst=upper
* compare
<1 aA
<3 aAt
<3 aa
<3 aat
<3 aaT
<3 ata
<3 aTa
** test: reset on expansion, ICU tickets 9415 & 9593
@ rules
&æ<x # tailor the last primary CE so that x sorts between ae and af
&æb=bæ # copy all reset CEs to make bæ sort the same
&각<h # copy/tailor 3 CEs to make h sort before the next Hangul syllable 갂
&⒀<<y # copy/tailor 4 CEs to make y sort with only a secondary difference
&l·=z # handle the pre-context for · when fetching reset CEs
<<u # copy/tailor 2 CEs
* compare
<1 ae
<2 æ
<1 x
<1 af
* compare
<1 aeb
<2 æb
= bæ
* compare
<1 각
<1 h
<1 갂
<1 갃
* compare
<1 · # by itself: primary CE
<1 l
<2 l· # l+middle dot has only a secondary difference from l
= z
<2 u
* compare
<1 (13)
<3 ⒀ # DUCET sets special tertiary weights in all CEs
<2 y
<1 (13[
% alternate=shifted
* compare
<1 (13)
= 13
<3 ⒀
= y # alternate=shifted removes the tailoring difference on the last CE
<1 14
** test: contraction inside extension, ICU ticket 9378
@ rules
&а<<х/й # all letters are Cyrillic
* compare
<1 ай
<2 х
** test: no duplicate tailored CEs for different reset positions with same CEs, ICU ticket 10104
@ rules
&t<x &ᵀ<y # same primary weights
&q<u &[before 1]ꝗ<v # q and ꝗ are primary adjacent
* compare
<1 q
<1 u
<1 v
<1 ꝗ
<1 t
<3 ᵀ
<1 y
<1 x
# Principle: Each rule builds on the state of preceding rules and ignores following rules.
** test: later rule does not affect earlier reset position, ICU ticket 10105
@ rules
&a < u < v < w &ov < x &b < v
* compare
<1 oa
<1 ou
<1 x # CE(o) followed by CE between u and w
<1 ow
<1 ob
<1 ov
** test: later rule does not affect earlier extension (1), ICU ticket 10105
@ rules
&a=x/b &v=b
% strength=secondary
* compare
<1 B
<1 c
<1 v
= b
* compare
<1 AB
= x
<1 ac
<1 av
= ab
** test: later rule does not affect earlier extension (2), ICU ticket 10105
@ rules
&a <<< c / e &g <<< e / l
% strength=secondary
* compare
<1 AE
= c
<2 æ
<1 agl
= ae
** test: later rule does not affect earlier extension (3), ICU ticket 10105
@ rules
&a = b / c &d = c / e
% strength=secondary
* compare
<1 AC # C is still only tertiary different from the original c
= b
<1 ade
= ac
** test: extension contains tailored character, ICU ticket 10105
@ rules
&a=e &b=u/e
* compare
<1 a
= e
<1 ba
= be
= u
** test: add simple mappings for characters with root context
@ rules
&z=· # middle dot has a prefix mapping in the CLDR root
&n=и # и (U+0438) has contractions in the root
* compare
<1 l
<2 l· # root mapping for l|· still works
<1 z
= ·
* compare
<1 n
= и
<1 И
<1 и\u0306 # root mapping for й=и\u0306 still works
= й
<3 Й
** test: add context mappings around characters with root context
@ rules
&z=·h # middle dot has a prefix mapping in the CLDR root
&n=ә|и # и (U+0438) has contractions in the root
* compare
<1 l
<2 l· # root mapping for l|· still works
<1 z
= ·h
* compare
<1 и
<3 И
<1 и\u0306 # root mapping for й=и\u0306 still works
= й
* compare
<1 әn
= әи
<1 әo
** test: many secondary CEs at the top of their range
@ rules
&[last primary ignorable]<<*\u2801-\u28ff
* compare
<2 \u0308
<2 \u2801
<2 \u2802
<2 \u2803
<2 \u2804
<2 \u28fd
<2 \u28fe
<2 \u28ff
<1 \x20
** test: many tertiary CEs at the top of their range
@ rules
&[last secondary ignorable]<<<*a-z
* compare
<3 a
<3 b
<3 c
<3 d
# e..w
<3 x
<3 y
<3 z
<2 \u0308
** test: tailor contraction together with nearly equivalent prefix, ICU ticket 10101
@ rules
&a=p|x &b=px &c=op
* compare
<1 b
= px
<3 B
<1 c
= op
<3 C
* compare
<1 ca
= opx # first contraction op, then prefix p|x
<3 cA
<3 Ca
** test: reset position with prefix (pre-context), ICU ticket 10102
@ rules
&a=p|x &px=y
* compare
<1 pa
= px
= y
<3 pA
<1 q
<1 x
** test: prefix+contraction together (1), ICU ticket 10071
@ rules
&x=a|bc
* compare
<1 ab
<1 Abc
<1 abd
<1 ac
<1 aw
<1 ax
= abc
<3 aX
<3 Ax
<1 b
<1 bb
<1 bc
<3 bC
<3 Bc
<1 bd
** test: prefix+contraction together (2), ICU ticket 10071
@ rules
&w=bc &x=a|b
* compare
<1 w
= bc
<3 W
* compare
<1 aw
<1 ax
= ab
<3 aX
<1 axb
<1 axc
= abc # prefix match a|b takes precedence over contraction match bc
<3 abC
<1 abd
<1 ay
** test: prefix+contraction together (3), ICU ticket 10071
@ rules
&x=a|b &w=bc # reverse order of rules as previous test, order should not matter here
* compare # same "compare" sequences as previous test
<1 w
= bc
<3 W
* compare
<1 aw
<1 ax
= ab
<3 aX
<1 axb
<1 axc
= abc # prefix match a|b takes precedence over contraction match bc
<3 abC
<1 abd
<1 ay
** test: no mapping p|c, falls back to contraction ch, CLDR ticket 5962
@ rules
&d=ch &v=p|ci
* compare
<1 pc
<3 pC
<1 pcH
<1 pcI
<1 pd
= pch # no-prefix contraction ch matches
<3 pD
<1 pv
= pci # prefix+contraction p|ci matches
<3 pV
** test: tailor in & around compact ranges of root primaries
# The Ogham characters U+1681..U+169A are in simple ascending order of primary CEs
# which should be reliably encoded as one range in the root elements data.
@ rules
&[before 1]ᚁ<a
&ᚁ<b
&[before 1]ᚂ<c
&ᚂ<d
&[before 1]ᚚ<y
&ᚚ<z
&[before 2]ᚁ<<r
&ᚁ<<s
&[before 3]ᚚ<<<t
&ᚚ<<<u
* compare
<1 ᣵ # U+18F5 last Canadian Aboriginal
<1 a
<1 r
<2 ᚁ
<2 s
<1 b
<1 c
<1 ᚂ
<1 d
<1 ᚃ
<1 ᚙ
<1 y
<1 t
<3 ᚚ
<3 u
<1 z
<1 ᚠ # U+16A0 first Runic
** test: suppressContractions
@ rules
&z<ch<әж [suppressContractions [·cә]]
* compare
<1 ch
<3 cH # ch was suppressed
<1 l
<1 l· # primary difference, not secondary, because l|· was suppressed
<1 ә
<2 ә\u0308 # secondary difference, not primary, because contractions for ә were suppressed
<1 әж
<3 әЖ
** test: Hangul & Jamo
@ rules
&L=\u1100 # first Jamo L
&V=\u1161 # first Jamo V
&T=\u11A8 # first Jamo T
&\uAC01<<*\u4E00-\u4EFF # first Hangul LVT syllable & lots of secondary diffs
* compare
<1 Lv
<3 LV
= \u1100\u1161
= \uAC00
<1 LVt
<3 LVT
= \u1100\u1161\u11A8
= \uAC00\u11A8
= \uAC01
<2 LVT\u0308
<2 \u4E00
<2 \u4E01
<2 \u4E80
<2 \u4EFF
<2 LV\u0308T
<1 \uAC02
** test: adjust special reset positions according to previous rules, CLDR ticket 6070
@ rules
&[last variable]<x
[maxVariable space] # has effect only after building, no effect on following rules
&[last variable]<y
&[before 1][first regular]<z
* compare
<1 ? # some punctuation
<1 x
<1 y
<1 z
<1 $ # some symbol
@ rules
&[last primary ignorable]<<x<<<y
&[last primary ignorable]<<z
* compare
<2 \u0358
<2 x
<3 y
<2 z
<1 \x20
@ rules
&[last secondary ignorable]<<<x
&[last secondary ignorable]<<<y
* compare
<3 x
<3 y
<2 \u0358
@ rules
&[before 2][first variable]<<z
&[before 2][first variable]<<y
&[before 3][first variable]<<<x
&[before 3][first variable]<<<w
&[before 1][first variable]<v
&[before 2][first variable]<<u
&[before 3][first variable]<<<t
&[before 2]\uFDD1\xA0<<s # FractionalUCA.txt: FDD1 00A0, SPACE first primary
* compare
<2 \u0358
<1 s
<2 \uFDD1\xA0
<1 t
<3 u
<2 v
<1 w
<3 x
<3 y
<2 z
<2 \t
@ rules
&[before 2][first regular]<<z
&[before 3][first regular]<<<y
&[before 1][first regular]<x
&[before 3][first regular]<<<w
&[before 2]\uFDD1\u263A<<v # FractionalUCA.txt: FDD1 263A, SYMBOL first primary
&[before 3][first regular]<<<u
&[before 1][first regular]<p # primary before the boundary: becomes variable
&[before 3][first regular]<<<t # not affected by p
&[last variable]<q # after p!
* compare
<1 ?
<1 p
<1 q
<1 t
<3 u
<3 v
<1 w
<3 x
<1 y
<3 z
<1 $
# check that p & q are indeed variable
% alternate=shifted
* compare
= ?
= p
= q
<1 t
<3 u
<3 v
<1 w
<3 x
<1 y
<3 z
<1 $
@ rules
&[before 2][first trailing]<<z
&[before 1][first trailing]<y
&[before 3][first trailing]<<<x
* compare
<1 \u4E00 # first Han, first implicit
<1 \uFDD1\uFDD0 # FractionalUCA.txt: unassigned first primary
# Note: The root collator currently does not map any characters to the trailing first boundary primary.
<1 x
<3 y
<1 z
<2 \uFFFD # The root collator currently maps U+FFFD to the first real trailing primary.
@ rules
&[before 2][first primary ignorable]<<z
&[before 2][first primary ignorable]<<y
&[before 3][first primary ignorable]<<<x
&[before 3][first primary ignorable]<<<w
* compare
= \x01
<2 w
<3 x
<3 y
<2 z
<2 \u0301
@ rules
&[before 3][first secondary ignorable]<<<y
&[before 3][first secondary ignorable]<<<x
* compare
= \x01
<3 x
<3 y
<2 \u0301
** test: canonical closure
@ rules
&X=A &U=Â
* compare
<1 U
= Â
= A\u0302
<2 Ú # U with acute
= U\u0301
= Ấ # A with circumflex & acute
= Â\u0301
= A\u0302\u0301
<1 X
= A
<2 X\u030A # with ring above
= Å
= A\u030A
= \u212B # Angstrom sign
@ rules
&x=\u5140\u55C0
* compare
<1 x
= \u5140\u55C0
= \u5140\uFA0D
= \uFA0C\u55C0
= \uFA0C\uFA0D # CJK compatibility characters
<3 X
# canonical closure on prefix rules, ICU ticket 9444
@ rules
&x=ä|ŝ
* compare
<1 äs # not tailored
<1 äx
= äŝ
= a\u0308s\u0302
= a\u0308ŝ
= äs\u0302
<3 äX
** test: conjoining Jamo map to expansions
@ rules
&gg=\u1101 # Jamo Lead consonant GG
&nj=\u11AC # Jamo Trail consonant NJ
* compare
<1 gg\u1161nj
= \u1101\u1161\u11AC
= \uAE4C\u11AC
= \uAE51
<3 gg\u1161nJ
<1 \u1100\u1100
** test: canonical tail closure, ICU ticket 5913
@ rules
&a<â
* compare
<1 a
<1 â # tailored
= a\u0302
<2 a\u0323\u0302 # discontiguous contraction
= ạ\u0302 # equivalent
= ậ # equivalent
<1 b
@ rules
&a<ạ
* compare
<1 a
<1 ạ # tailored
= a\u0323
<2 a\u0323\u0302 # contiguous contraction plus extra diacritic
= ạ\u0302 # equivalent
= ậ # equivalent
<1 b
# Tail closure should work even if there is a prefix and/or contraction.
@ rules
&a<\u5140|câ
# In order to find discontiguous contractions for \u5140|câ
# there must exist a mapping for \u5140|ca, regardless of what it maps to.
# (This follows from the UCA spec.)
&x=\u5140|ca
* compare
<1 \u5140a
= \uFA0Ca
<1 \u5140câ # tailored
= \uFA0Ccâ
= \u5140ca\u0302
= \uFA0Cca\u0302
<2 \u5140ca\u0323\u0302 # discontiguous contraction
= \uFA0Cca\u0323\u0302
= \u5140cạ\u0302
= \uFA0Ccạ\u0302
= \u5140cậ
= \uFA0Ccậ
<1 \u5140b
= \uFA0Cb
<1 \u5140x
= \u5140ca
# Double-check that without the extra mapping there will be no discontiguous match.
@ rules
&a<\u5140|câ
* compare
<1 \u5140a
= \uFA0Ca
<1 \u5140câ # tailored
= \uFA0Ccâ
= \u5140ca\u0302
= \uFA0Cca\u0302
<1 \u5140b
= \uFA0Cb
<1 \u5140ca\u0323\u0302 # no discontiguous contraction
= \uFA0Cca\u0323\u0302
= \u5140cạ\u0302
= \uFA0Ccạ\u0302
= \u5140cậ
= \uFA0Ccậ
@ rules
&a<cạ
* compare
<1 a
<1 cạ # tailored
= ca\u0323
<2 ca\u0323\u0302 # contiguous contraction plus extra diacritic
= cạ\u0302 # equivalent
= cậ # equivalent
<1 b
# ᾢ = U+1FA2 GREEK SMALL LETTER OMEGA WITH PSILI AND VARIA AND YPOGEGRAMMENI
# = 03C9 0313 0300 0345
# ccc = 0, 230, 230, 240
@ rules
&δ=αῳ
# In order to find discontiguous contractions for αῳ
# there must exist a mapping for αω, regardless of what it maps to.
# (This follows from the UCA spec.)
&ε=αω
* compare
<1 δ
= αῳ
= αω\u0345
<2 αω\u0313\u0300\u0345 # discontiguous contraction
= αὠ\u0300\u0345
= αὢ\u0345
= αᾢ
<2 αω\u0300\u0313\u0345
= αὼ\u0313\u0345
= αῲ\u0313 # not FCD
<1 ε
= αω
# Double-check that without the extra mapping there will be no discontiguous match.
@ rules
&δ=αῳ
* compare
<1 αω\u0313\u0300\u0345 # no discontiguous contraction
= αὠ\u0300\u0345
= αὢ\u0345
= αᾢ
<2 αω\u0300\u0313\u0345
= αὼ\u0313\u0345
= αῲ\u0313 # not FCD
<1 δ
= αῳ
= αω\u0345
# Add U+0315 COMBINING COMMA ABOVE RIGHT which has ccc=232.
# Tests code paths where the tailored string has a combining mark
# that does not occur in any composite's decomposition.
@ rules
&δ=αὼ\u0315
* compare
<1 αω\u0313\u0300\u0315 # Not tailored: The grave accent blocks the comma above.
= αὠ\u0300\u0315
= αὢ\u0315
<1 δ
= αὼ\u0315
= αω\u0300\u0315
<2 αω\u0300\u0315\u0345
= αὼ\u0315\u0345
= αῲ\u0315 # not FCD
** test: danish a+a vs. a-umlaut, ICU ticket 9319
@ rules
&z<aa
* compare
<1 z
<1 aa
<2 aa\u0308
= aä
** test: Jamo L with and in prefix
# Useful for the Korean "searchjl" tailoring (instead of contractions of pairs of Jamo L).
@ rules
# Jamo Lead consonant G after G or GG
&[last primary ignorable]<<\u1100|\u1100=\u1101|\u1100
# Jamo Lead consonant GG sorts like G+G
&\u1100\u1100=\u1101
# Note: Making G|GG and GG|GG sort the same as G|G+G
# would require the ability to reset on G|G+G,
# or we could make G-after-G equal to some secondary-CE character,
# and reset on a pair of those.
# (It does not matter much if there are at most two G in a row in real text.)
* compare
<1 \u1100
<2 \u1100\u1100 # only one primary from a sequence of G lead consonants
= \u1101
<2 \u1100\u1100\u1100
= \u1101\u1100
# but not = \u1100\u1101, see above
<1 \u1100\u1161
= \uAC00
<2 \u1100\u1100\u1161
= \u1100\uAC00 # prefix match from the L of the LV syllable
= \u1101\u1161
= \uAE4C
** test: proposed Korean "searchjl" tailoring with prefixes, CLDR ticket 6546
@ rules
# Low secondary CEs for Jamo V & T.
# Note: T should sort before V for proper syllable order.
&\u0332 # COMBINING LOW LINE (first primary ignorable)
<<\u1161<<\u1162
# Korean Jamo lead consonant search rules, part 2:
# Make modern compound L jamo primary equivalent to non-compound forms.
# Secondary CEs for Jamo L-after-L, greater than Jamo V & T.
&\u0313 # COMBINING COMMA ABOVE (second primary ignorable)
=\u1100|\u1100
=\u1103|\u1103
=\u1107|\u1107
=\u1109|\u1109
=\u110C|\u110C
# Compound L Jamo map to equivalent expansions of primary+secondary CE.
&\u1100\u0313=\u1101<<<\u3132 # HANGUL CHOSEONG SSANGKIYEOK, HANGUL LETTER SSANGKIYEOK
&\u1103\u0313=\u1104<<<\u3138 # HANGUL CHOSEONG SSANGTIKEUT, HANGUL LETTER SSANGTIKEUT
&\u1107\u0313=\u1108<<<\u3143 # HANGUL CHOSEONG SSANGPIEUP, HANGUL LETTER SSANGPIEUP
&\u1109\u0313=\u110A<<<\u3146 # HANGUL CHOSEONG SSANGSIOS, HANGUL LETTER SSANGSIOS
&\u110C\u0313=\u110D<<<\u3149 # HANGUL CHOSEONG SSANGCIEUC, HANGUL LETTER SSANGCIEUC
* compare
<1 \u1100\u1161
= \uAC00
<2 \u1100\u1162
= \uAC1C
<2 \u1100\u1100\u1161
= \u1100\uAC00
= \u1101\u1161
= \uAE4C
<3 \u3132\u1161
** test: Hangul syllables in prefix & in the interior of a contraction
@ rules
&x=\u1100\u1161|a\u1102\u1162z
* compare
<1 \u1100\u1161x
= \u1100\u1161a\u1102\u1162z
= \u1100\u1161a\uB0B4z
= \uAC00a\u1102\u1162z
= \uAC00a\uB0B4z
** test: digits are unsafe-backwards when numeric=on
@ root
% numeric=on
* compare
# If digits are not unsafe, then numeric collation sees "1"=="01" and "b">"a".
# We need to back up before the identical prefix "1" and compare the full numbers.
<1 11b
<1 101a
** test: simple locale data test
@ locale de
* compare
<1 a
<2 ä
<1 ae
<2 æ
@ locale de-u-co-phonebk
* compare
<1 a
<1 ae
<2 ä
<2 æ
# The following test cases were moved here from ICU 52's DataDrivenCollationTest.txt.
** test: DataDrivenCollationTest/TestMorePinyin
# Testing the primary strength.
@ locale zh
% strength=primary
* compare
< lā
= lĀ
= Lā
= LĀ
< lān
= lĀn
< lē
= lĒ
= Lē
= LĒ
< lēn
= lĒn
** test: DataDrivenCollationTest/TestLithuanian
# Lithuanian sort order.
@ locale lt
* compare
< cz
< č
< d
< iz
< j
< sz
< š
< t
< zz
< ž
** test: DataDrivenCollationTest/TestLatvian
# Latvian sort order.
@ locale lv
* compare
< cz
< č
< d
< gz
< ģ
< h
< iz
< j
< kz
< ķ
< l
< lz
< ļ
< m
< nz
< ņ
< o
< rz
< ŗ
< s
< sz
< š
< t
< zz
< ž
** test: DataDrivenCollationTest/TestEstonian
# Estonian sort order.
@ locale et
* compare
< sy
< š
< šy
< z
< zy
< ž
< v
< va
< w
< õ
< õy
< ä
< äy
< ö
< öy
< ü
< üy
< x
** test: DataDrivenCollationTest/TestAlbanian
# Albanian sort order.
@ locale sq
* compare
< cz
< ç
< d
< dz
< dh
< e
< ez
< ë
< f
< gz
< gj
< h
< lz
< ll
< m
< nz
< nj
< o
< rz
< rr
< s
< sz
< sh
< t
< tz
< th
< u
< xz
< xh
< y
< zz
< zh
** test: DataDrivenCollationTest/TestSimplifiedChineseOrder
# Sorted file has different order.
@ root
# normalization=on turned on & off automatically.
* compare
< \u5F20
< \u5F20\u4E00\u8E3F
** test: DataDrivenCollationTest/TestTibetanNormalizedIterativeCrash
# This pretty much crashes.
@ root
* compare
< \u0f71\u0f72\u0f80\u0f71\u0f72
< \u0f80
** test: DataDrivenCollationTest/TestThaiPartialSortKeyProblems
# These are examples of strings that caused trouble in partial sort key testing.
@ locale th-TH
* compare
< \u0E01\u0E01\u0E38\u0E18\u0E20\u0E31\u0E13\u0E11\u0E4C
< \u0E01\u0E01\u0E38\u0E2A\u0E31\u0E19\u0E42\u0E18
* compare
< \u0E01\u0E07\u0E01\u0E32\u0E23
< \u0E01\u0E07\u0E42\u0E01\u0E49
* compare
< \u0E01\u0E23\u0E19\u0E17\u0E32
< \u0E01\u0E23\u0E19\u0E19\u0E40\u0E0A\u0E49\u0E32
* compare
< \u0E01\u0E23\u0E30\u0E40\u0E08\u0E35\u0E22\u0E27
< \u0E01\u0E23\u0E30\u0E40\u0E08\u0E35\u0E4A\u0E22\u0E27
* compare
< \u0E01\u0E23\u0E23\u0E40\u0E0A\u0E2D
< \u0E01\u0E23\u0E23\u0E40\u0E0A\u0E49\u0E32
** test: DataDrivenCollationTest/TestJavaStyleRule
# java.text allows rules to start as '<<<x<<<y...'
# we emulate this by assuming a &[first tertiary ignorable] in this case.
@ rules
&\u0001=equal<<<z<<x<<<w &[first tertiary ignorable]=a &[first primary ignorable]=b
* compare
= a
= equal
< z
< x
= b # x had become the new first primary ignorable
< w
** test: DataDrivenCollationTest/TestShiftedIgnorable
# The UCA states that primary ignorables should be completely
# ignorable when following a shifted code point.
@ root
% alternate=shifted
% strength=quaternary
* compare
< a\u0020b
= a\u0020\u0300b
= a\u0020\u0301b
< a_b
= a_\u0300b
= a_\u0301b
< A\u0020b
= A\u0020\u0300b
= A\u0020\u0301b
< A_b
= A_\u0300b
= A_\u0301b
< a\u0301b
< A\u0301b
< a\u0300b
< A\u0300b
** test: DataDrivenCollationTest/TestNShiftedIgnorable
# The UCA states that primary ignorables should be completely
# ignorable when following a shifted code point.
@ root
% alternate=non-ignorable
% strength=tertiary
* compare
< a\u0020b
< A\u0020b
< a\u0020\u0301b
< A\u0020\u0301b
< a\u0020\u0300b
< A\u0020\u0300b
< a_b
< A_b
< a_\u0301b
< A_\u0301b
< a_\u0300b
< A_\u0300b
< a\u0301b
< A\u0301b
< a\u0300b
< A\u0300b
** test: DataDrivenCollationTest/TestSafeSurrogates
# It turned out that surrogates were not skipped properly
# when iterating backwards if they were in the middle of a
# contraction. This test assures that this is fixed.
@ rules
&a < x\ud800\udc00b
* compare
< a
< x\ud800\udc00b
** test: DataDrivenCollationTest/da_TestPrimary
# This test goes through primary strength cases
@ locale da
% strength=primary
* compare
< Lvi
< Lwi
* compare
< L\u00e4vi
< L\u00f6wi
* compare
< L\u00fcbeck
= Lybeck
** test: DataDrivenCollationTest/da_TestTertiary
# This test goes through tertiary strength cases
@ locale da
% strength=tertiary
* compare
< Luc
< luck
* compare
< luck
< L\u00fcbeck
* compare
< lybeck
< L\u00fcbeck
* compare
< L\u00e4vi
< L\u00f6we
* compare
< L\u00f6ww
< mast
* compare
< A/S
< ANDRE
< ANDR\u00c9
< ANDREAS
< AS
< CA
< \u00c7A
< CB
< \u00c7C
< D.S.B.
< DA
< \u00d0A
< DB
< \u00d0C
< DSB
< DSC
< EKSTRA_ARBEJDE
< EKSTRABUD0
< H\u00d8ST
< HAAG
< H\u00c5NDBOG
< HAANDV\u00c6RKSBANKEN
< Karl
< karl
< NIELS\u0020J\u00d8RGEN
< NIELS-J\u00d8RGEN
< NIELSEN
< R\u00c9E,\u0020A
< REE,\u0020B
< R\u00c9E,\u0020L
< REE,\u0020V
< SCHYTT,\u0020B
< SCHYTT,\u0020H
< SCH\u00dcTT,\u0020H
< SCHYTT,\u0020L
< SCH\u00dcTT,\u0020M
< SS
< \u00df
< SSA
< STORE\u0020VILDMOSE
< STOREK\u00c6R0
< STORM\u0020PETERSEN
< STORMLY
< THORVALD
< THORVARDUR
< \u00feORVAR\u00d0UR
< THYGESEN
< VESTERG\u00c5RD,\u0020A
< VESTERGAARD,\u0020A
< VESTERG\u00c5RD,\u0020B
< \u00c6BLE
< \u00c4BLE
< \u00d8BERG
< \u00d6BERG
* compare
< andere
< chaque
< chemin
< cote
< cot\u00e9
< c\u00f4te
< c\u00f4t\u00e9
< \u010du\u010d\u0113t
< Czech
< hi\u0161a
< irdisch
< lie
< lire
< llama
< l\u00f5ug
< l\u00f2za
< lu\u010d
< luck
< L\u00fcbeck
< lye
< l\u00e4vi
< L\u00f6wen
< m\u00e0\u0161ta
< m\u00eer
< myndig
< M\u00e4nner
< m\u00f6chten
< pi\u00f1a
< pint
< pylon
< \u0161\u00e0ran
< savoir
< \u0160erb\u016bra
< Sietla
< \u015blub
< subtle
< symbol
< s\u00e4mtlich
< verkehrt
< vox
< v\u00e4ga
< waffle
< wood
< yen
< yuan
< yucca
< \u017eal
< \u017eena
< \u017den\u0113va
< zoo0
< Zviedrija
< Z\u00fcrich
< zysk0
< \u00e4ndere
** test: DataDrivenCollationTest/hi_TestNewRules
# This test goes through new rules and tests against old rules
@ locale hi
* compare
< कॐ
< कं
< कँ
< कः
** test: DataDrivenCollationTest/ro_TestNewRules
# This test goes through new rules and tests against old rules
@ locale ro
* compare
< xAx
< xă
< xĂ
< Xă
< XĂ
< xăx
< xĂx
< xâ
< xÂ
< Xâ
< XÂ
< xâx
< xÂx
< xb
< xIx
< xî
< xÎ
< Xî
< XÎ
< xîx
< xÎx
< xj
< xSx
< xș
= xş
< xȘ
= xŞ
< Xș
= Xş
< XȘ
= XŞ
< xșx
= xşx
< xȘx
= xŞx
< xT
< xTx
< xț
= xţ
< xȚ
= xŢ
< Xț
= Xţ
< XȚ
= XŢ
< xțx
= xţx
< xȚx
= xŢx
< xU
** test: DataDrivenCollationTest/testOffsets
# This tests cases where forwards and backwards iteration get different offsets
@ locale en
% strength=tertiary
* compare
< a\uD800\uDC00\uDC00
< b\uD800\uDC00\uDC00
* compare
< \u0301A\u0301\u0301
< \u0301B\u0301\u0301
* compare
< abcd\r\u0301
< abce\r\u0301
# TODO: test offsets in new CollationTest
# End of test cases moved here from ICU 52's DataDrivenCollationTest.txt.
** test: was ICU 52 cmsccoll/TestRedundantRules
@ rules
& a < b < c < d& [before 1] c < m
* compare
<1 a
<1 b
<1 m
<1 c
<1 d
@ rules
& a < b <<< c << d <<< e& [before 3] e <<< x
* compare
<1 a
<1 b
<3 c
<2 d
<3 x
<3 e
@ rules
& a < b <<< c << d <<< e <<< f < g& [before 1] g < x
* compare
<1 a
<1 b
<3 c
<2 d
<3 e
<3 f
<1 x
<1 g
@ rules
& a <<< b << c < d& a < m
* compare
<1 a
<3 b
<2 c
<1 m
<1 d
@ rules
&a<b<<b\u0301 &z<b
* compare
<1 a
<1 b\u0301
<1 z
<1 b
@ rules
&z<m<<<q<<<m
* compare
<1 z
<1 q
<3 m
@ rules
&z<<<m<q<<<m
* compare
<1 z
<1 q
<3 m
@ rules
& a < b < c < d& r < c
* compare
<1 a
<1 b
<1 d
<1 r
<1 c
@ rules
& a < b < c < d& c < m
* compare
<1 a
<1 b
<1 c
<1 m
<1 d
@ rules
& a < b < c < d& a < m
* compare
<1 a
<1 m
<1 b
<1 c
<1 d
** test: was ICU 52 cmsccoll/TestExpansionSyntax
# The following two rules should sort the particular list of strings the same.
@ rules
&AE <<< a << b <<< c &d <<< f
* compare
<1 AE
<3 a
<2 b
<3 c
<1 d
<3 f
@ rules
&A <<< a / E << b / E <<< c /E &d <<< f
* compare
<1 AE
<3 a
<2 b
<3 c
<1 d
<3 f
# The following two rules should sort the particular list of strings the same.
@ rules
&AE <<< a <<< b << c << d < e < f <<< g
* compare
<1 AE
<3 a
<3 b
<2 c
<2 d
<1 e
<1 f
<3 g
@ rules
&A <<< a / E <<< b / E << c / E << d / E < e < f <<< g
* compare
<1 AE
<3 a
<3 b
<2 c
<2 d
<1 e
<1 f
<3 g
# The following two rules should sort the particular list of strings the same.
@ rules
&AE <<< B <<< C / D <<< F
* compare
<1 AE
<3 B
<3 F
<1 AED
<3 C
@ rules
&A <<< B / E <<< C / ED <<< F / E
* compare
<1 AE
<3 B
<3 F
<1 AED
<3 C
** test: never reorder trailing primaries
@ root
% reorder Zzzz Grek
* compare
<1 L
<1 字
<1 Ω
<1 \uFFFD
<1 \uFFFF
** test: fall back to mappings with shorter prefixes, not immediately to ones with no prefixes
@ rules
&u=ab|cd
&v=b|ce
* compare
<1 abc
<1 abcc
<1 abcf
<1 abcd
= abu
<1 abce
= abv
# With the following rules, there is only one prefix per composite ĉ or ç,
# but both prefixes apply to just c in NFD form.
# We would get different results for composed vs. NFD input
# if we fell back directly from longest-prefix mappings to no-prefix mappings.
@ rules
&x=op|ĉ
&y=p|ç
* compare
<1 opc
<2 opć
<1 opcz
<1 opd
<1 opĉ
= opc\u0302
= opx
<1 opç
= opc\u0327
= opy
# The mapping is used which has the longest matching prefix for which
# there is also a suffix match, with the longest suffix match among several for that prefix.
@ rules
&❶=d
&❷=de
&❸=def
&①=c|d
&②=c|de
&③=c|def
&④=bc|d
&⑤=bc|de
&⑥=bc|def
&⑦=abc|d
&⑧=abc|de
&⑨=abc|def
* compare
<1 9aadzz
= 9aa❶zz
<1 9aadez
= 9aa❷z
<1 9aadef
= 9aa❸
<1 9acdzz
= 9ac①zz
<1 9acdez
= 9ac②z
<1 9acdef
= 9ac③
<1 9bcdzz
= 9bc④zz
<1 9bcdez
= 9bc⑤z
<1 9bcdef
= 9bc⑥
<1 abcdzz
= abc⑦zz
<1 abcdez
= abc⑧z
<1 abcdef
= abc⑨
** test: prefix + discontiguous contraction with missing prefix contraction
# Unfortunate terminology: The first "prefix" here is the pre-context,
# the second "prefix" refers to the contraction/relation string that is
# one shorter than the one being tested.
@ rules
&x=p|e
&y=p|ê
&z=op|ê
# No mapping for op|e:
# Discontiguous contraction matching should not match op|ê in opệ
# because it would have to skip the dot below and extend a match on op|e by the circumflex,
# but there is no match on op|e.
* compare
<1 oPe
<1 ope
= opx
<1 opệ
= opy\u0323 # y not z
<1 opê
= opz
# We cannot test for fallback by whether the contraction default CE32
# is for another contraction. With the following rules, there is no mapping for op|e,
# and the fallback to prefix p has no contractions.
@ rules
&x=p|e
&z=op|ê
* compare
<1 oPe
<1 ope
= opx
<2 opệ
= opx\u0323\u0302 # x not z
<1 opê
= opz
# One more variation: Fallback to the simple code point, no shorter non-empty prefix.
@ rules
&x=e
&z=op|ê
* compare
<1 ope
= opx
<3 oPe
= oPx
<2 opệ
= opx\u0323\u0302 # x not z
<1 opê
= opz
** test: maxVariable via rules
@ rules
[maxVariable space][alternate shifted]
* compare
= \u0020
= \u000A
<1 .
<1 ° # degree sign
<1 $
<1 0
** test: maxVariable via setting
@ root
% maxVariable=currency
% alternate=shifted
* compare
= \u0020
= \u000A
= .
= ° # degree sign
= $
<1 0
** test: ICU4J CollationMiscTest/TestContractionClosure (ää)
# This tests canonical closure, but it also tests that CollationFastLatin
# bails out properly for contractions with combining marks.
# For that we need pairs of strings that remain in the Latin fastpath
# long enough, hence the extra "= b" lines.
@ rules
&b=\u00e4\u00e4
* compare
<1 b
= \u00e4\u00e4
= b
= a\u0308a\u0308
= b
= \u00e4a\u0308
= b
= a\u0308\u00e4
** test: ICU4J CollationMiscTest/TestContractionClosure (Å)
@ rules
&b=\u00C5
* compare
<1 b
= \u00C5
= b
= A\u030A
= b
= \u212B
** test: reset-before on already-tailored characters, ICU ticket 10108
@ rules
&a<w<<x &[before 2]x<<y
* compare
<1 a
<1 w
<2 y
<2 x
@ rules
&a<<w<<<x &[before 2]x<<y
* compare
<1 a
<2 y
<2 w
<3 x
@ rules
&a<w<x &[before 2]x<<y
* compare
<1 a
<1 w
<1 y
<2 x
@ rules
&a<w<<<x &[before 2]x<<y
* compare
<1 a
<1 y
<2 w
<3 x
** test: numeric collation with other settings, ICU ticket 9092
@ root
% strength=identical
% caseFirst=upper
% numeric=on
* compare
<1 100\u0020a
<1 101
** test: collation type fallback from unsupported type, ICU ticket 10149
@ locale fr-CA-u-co-phonebk
# Expect the same result as with fr-CA, using backwards-secondary order.
# That is, we should fall back from the unsupported collation type
# to the locale's default collation type.
* compare
<1 cote
<2 côte
<2 coté
<2 côté
** test: @ is equivalent to [backwards 2], ICU ticket 9956
@ rules
&b<a @ &v<<w
* compare
<1 b
<1 a
<1 cote
<2 côte
<2 coté
<2 côté
<1 v
<2 w
<1 x
** test: shifted+reordering, ICU ticket 9507
@ root
% reorder Grek punct space
% alternate=shifted
% strength=quaternary
# Which primaries are "variable" should be determined without script reordering,
# and then primaries should be reordered whether they are shifted to quaternary or not.
* compare
<4 ( # punctuation
<4 )
<4 \u0020 # space
<1 ` # symbol
<1 ^
<1 $ # currency symbol
<1 €
<1 0 # numbers
<1 ε # Greek
<1 e # Latin
<1 e(e
<4 e)e
<4 e\u0020e
<4 ee
<3 e(E
<4 e)E
<4 e\u0020E
<4 eE
** test: "uppercase first" could sort a string before its prefix, ICU ticket 9351
@ rules
&\u0001<<<b<<<B
% caseFirst=upper
* compare
<1 aaa
<3 aaaB
** test: secondary+case ignores secondary ignorables, ICU ticket 9355
@ rules
&\u0001<<<b<<<B
% strength=secondary
% caseLevel=on
* compare
<1 a
= ab
= aB
** test: custom collation rules involving tail of a contraction in Malayalam, ICU ticket 6328
@ rules
&[before 2] ൌ << ൗ # U+0D57 << U+0D4C == 0D46+0D57
* compare
<1 ൗx
<2 ൌx
<1 ൗy
<2 ൌy
** test: quoted apostrophe in compact syntax, ICU ticket 8204
@ rules
&q<<*a''c
* compare
<1 d
<1 p
<1 q
<2 a
<2 \u0027
<2 c
<1 r
# ICU ticket #8260 "Support all collation-related keywords in Collator.getInstance()"
** test: locale -u- with collation keywords, ICU ticket 8260
@ locale de-u-kv-sPace-ka-shifTed-kn-kk-falsE-kf-Upper-kc-tRue-ks-leVel4
* compare
<4 \u0020 # space is shifted, strength=quaternary
<1 ! # punctuation is regular
<1 2
<1 12 # numeric sorting
<1 B
<c b # uppercase first on case level
<1 x\u0301\u0308
<2 x\u0308\u0301 # normalization off
** test: locale @ with collation keywords, ICU ticket 8260
@ locale fr@colbAckwards=yes;ColStrength=Quaternary;kv=currencY;colalternate=shifted
* compare
<4 $ # currency symbols are shifted, strength=quaternary
<1 àla
<2 alà # backwards secondary level
** test: locale -u- with script reordering, ICU ticket 8260
@ locale el-u-kr-kana-SYMBOL-Grek-hani-cyrl-latn-digit-armn-deva-ethi-thai
* compare
<1 \u0020
<1 あ
<1 ☂
<1 Ω
<1 丂
<1 ж
<1 L
<1 4
<1 Ձ
<1 अ
<1 ሄ
<1 ฉ
** test: locale @collation=type should be case-insensitive
@ locale de@coLLation=PhoneBook
* compare
<1 ae
<2 ä
<3 Ä
** test: import root search rules plus German phonebook rules, ICU ticket 8962
@ locale de-u-co-search
* compare
<1 =
<1 ≠
<1 a
<1 ae
<2 ä
# Once more, but with runtime builder.
@ rules
[import und-u-co-search][import de-u-co-phonebk]
* compare
<1 =
<1 ≠
<1 a
<1 ae
<2 ä
# Once again, with import from "root" not "und" (as in a proper language tag).
@ rules
[import root-u-co-search][import de-u-co-phonebk]
* compare
<1 =
<1 ≠
<1 a
<1 ae
<2 ä
** test: import rules from a language with non-Latin native script, and reset the reordering, ICU ticket 10998
# Greek should sort Greek first.
@ rules
[import el]
* compare
<1 4
<1 Ω
<1 L
# Import Greek, and then reset the reordering.
@ rules
[import el][reorder Zzzz]
* compare
<1 4
<1 L
<1 Ω
# "others" is a synonym for Zzzz.
@ rules
[import el][reorder others]
* compare
<1 4
<1 L
<1 Ω
** test: regression test for CollationFastLatinBuilder, ICU ticket 11388
@ rules
&x<<aa<<<Aa<<<AA
% strength=secondary
* compare
<1 AA
<2 Aẩ
<2 aą
* compare
<1 AA
<2 aą
** test: tailor tertiary-after a common tertiary where there is a lower one
# Assume that Hiragana small A has a below-common tertiary, and Hiragana A has a common one.
# See ICU ticket 11448 & CLDR ticket 7222.
@ rules
&あ<<<x<<<y<<<z
* compare
<1 ぁ
<3 あ
<3 x
<3 y
<3 z
<3 ァ
<1 い
** test: tailor tertiary-after a below-common tertiary
@ rules
&ぁ<<<x<<<y<<<z
* compare
<1 ぁ
<3 x
<3 y
<3 z
<3 あ
<3 ァ
<1 い
** test: tailor tertiary-before a common tertiary where there is a lower one
@ rules
&[before 3]あ<<<x<<<y<<<z
* compare
<1 ぁ
<3 x
<3 y
<3 z
<3 あ
<3 ァ
<1 い
** test: tailor tertiary-before a below-common tertiary
@ rules
&[before 3]ぁ<<<x<<<y<<<z
* compare
<1 x
<3 y
<3 z
<3 ぁ
<3 あ
<3 ァ
<1 い
** test: reorder single scripts not groups, ICU ticket 11449
@ root
% reorder Goth Latn
* compare
<1 4
<1 𐌰 # Gothic
<1 L
<1 Ω
# Before ICU 55, the following reordered together with Gothic.
<1 𐌈 # Old Italic
<1 𐑐 # Shavian