Delete source/test

source/test/testdata/collationtest.txt

Index: source/test/testdata/collationtest.txt
diff --git a/source/test/testdata/collationtest.txt b/source/test/testdata/collationtest.txt
deleted file mode 100644
index 3a703cb10b6a091cb39eb33b96a4ce7b18d0ae53..0000000000000000000000000000000000000000
--- a/source/test/testdata/collationtest.txt
+++ /dev/null
@@ -1,2540 +0,0 @@
-# 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