diff options
author | Rob Pike <r@golang.org> | 2009-05-05 17:05:39 -0700 |
---|---|---|
committer | Rob Pike <r@golang.org> | 2009-05-05 17:05:39 -0700 |
commit | b6708b2dad4af458e69fde7fc596ecdce03280d0 (patch) | |
tree | 11cfc1b980ecce9cb0eec06f49cc53ce05faf350 /src/lib/utf8_test.go | |
parent | 658b10477a16151de7e4d51657993f95445f11ee (diff) | |
download | golang-b6708b2dad4af458e69fde7fc596ecdce03280d0.tar.gz |
directory-per-package step 1: move files from lib/X.go to lib/X/X.go
no substantive changes except:
- new Makefiles, all auto-generated
- go/src/lib/Makefile has been extensively edited
R=rsc
OCL=28310
CL=28310
Diffstat (limited to 'src/lib/utf8_test.go')
-rw-r--r-- | src/lib/utf8_test.go | 179 |
1 files changed, 0 insertions, 179 deletions
diff --git a/src/lib/utf8_test.go b/src/lib/utf8_test.go deleted file mode 100644 index 3ba5ee2b8..000000000 --- a/src/lib/utf8_test.go +++ /dev/null @@ -1,179 +0,0 @@ -// Copyright 2009 The Go Authors. All rights reserved. -// Use of this source code is governed by a BSD-style -// license that can be found in the LICENSE file. - -package utf8 - -import ( - "fmt"; - "io"; - "testing"; - "utf8"; -) - -type Utf8Map struct { - rune int; - str string; -} - -var utf8map = []Utf8Map { - Utf8Map{ 0x0000, "\x00" }, - Utf8Map{ 0x0001, "\x01" }, - Utf8Map{ 0x007e, "\x7e" }, - Utf8Map{ 0x007f, "\x7f" }, - Utf8Map{ 0x0080, "\xc2\x80" }, - Utf8Map{ 0x0081, "\xc2\x81" }, - Utf8Map{ 0x00bf, "\xc2\xbf" }, - Utf8Map{ 0x00c0, "\xc3\x80" }, - Utf8Map{ 0x00c1, "\xc3\x81" }, - Utf8Map{ 0x00c8, "\xc3\x88" }, - Utf8Map{ 0x00d0, "\xc3\x90" }, - Utf8Map{ 0x00e0, "\xc3\xa0" }, - Utf8Map{ 0x00f0, "\xc3\xb0" }, - Utf8Map{ 0x00f8, "\xc3\xb8" }, - Utf8Map{ 0x00ff, "\xc3\xbf" }, - Utf8Map{ 0x0100, "\xc4\x80" }, - Utf8Map{ 0x07ff, "\xdf\xbf" }, - Utf8Map{ 0x0800, "\xe0\xa0\x80" }, - Utf8Map{ 0x0801, "\xe0\xa0\x81" }, - Utf8Map{ 0xfffe, "\xef\xbf\xbe" }, - Utf8Map{ 0xffff, "\xef\xbf\xbf" }, - Utf8Map{ 0x10000, "\xf0\x90\x80\x80" }, - Utf8Map{ 0x10001, "\xf0\x90\x80\x81" }, - Utf8Map{ 0x10fffe, "\xf4\x8f\xbf\xbe" }, - Utf8Map{ 0x10ffff, "\xf4\x8f\xbf\xbf" }, -} - -// io.StringBytes with one extra byte at end -func bytes(s string) []byte { - s += "\x00"; - b := io.StringBytes(s); - return b[0:len(s)-1]; -} - -func TestFullRune(t *testing.T) { - for i := 0; i < len(utf8map); i++ { - m := utf8map[i]; - b := bytes(m.str); - if !utf8.FullRune(b) { - t.Errorf("FullRune(%q) (rune %04x) = false, want true", b, m.rune); - } - s := "xx"+m.str; - if !utf8.FullRuneInString(s, 2) { - t.Errorf("FullRuneInString(%q, 2) (rune %04x) = false, want true", s, m.rune); - } - b1 := b[0:len(b)-1]; - if utf8.FullRune(b1) { - t.Errorf("FullRune(%q) = true, want false", b1); - } - s1 := "xxx"+string(b1); - if utf8.FullRuneInString(s1, 3) { - t.Errorf("FullRune(%q, 3) = true, want false", s1); - } - } -} - -func equalBytes(a, b []byte) bool { - if len(a) != len(b) { - return false; - } - for i := 0; i < len(a); i++ { - if a[i] != b[i] { - return false; - } - } - return true; -} - -func TestEncodeRune(t *testing.T) { - for i := 0; i < len(utf8map); i++ { - m := utf8map[i]; - b := bytes(m.str); - var buf [10]byte; - n := utf8.EncodeRune(m.rune, &buf); - b1 := buf[0:n]; - if !equalBytes(b, b1) { - t.Errorf("EncodeRune(0x%04x) = %q want %q", m.rune, b1, b); - } - } -} - -func TestDecodeRune(t *testing.T) { - for i := 0; i < len(utf8map); i++ { - m := utf8map[i]; - b := bytes(m.str); - rune, size := utf8.DecodeRune(b); - if rune != m.rune || size != len(b) { - t.Errorf("DecodeRune(%q) = 0x%04x, %d want 0x%04x, %d", b, rune, size, m.rune, len(b)); - } - s := "xx"+m.str; - rune, size = utf8.DecodeRuneInString(s, 2); - if rune != m.rune || size != len(b) { - t.Errorf("DecodeRune(%q, 2) = 0x%04x, %d want 0x%04x, %d", s, rune, size, m.rune, len(b)); - } - - // there's an extra byte that bytes left behind - make sure trailing byte works - rune, size = utf8.DecodeRune(b[0:cap(b)]); - if rune != m.rune || size != len(b) { - t.Errorf("DecodeRune(%q) = 0x%04x, %d want 0x%04x, %d", b, rune, size, m.rune, len(b)); - } - s = "x"+m.str+"\x00"; - rune, size = utf8.DecodeRuneInString(s, 1); - if rune != m.rune || size != len(b) { - t.Errorf("DecodeRuneInString(%q, 1) = 0x%04x, %d want 0x%04x, %d", s, rune, size, m.rune, len(b)); - } - - // make sure missing bytes fail - wantsize := 1; - if wantsize >= len(b) { - wantsize = 0; - } - rune, size = utf8.DecodeRune(b[0:len(b)-1]); - if rune != RuneError || size != wantsize { - t.Errorf("DecodeRune(%q) = 0x%04x, %d want 0x%04x, %d", b[0:len(b)-1], rune, size, RuneError, wantsize); - } - s = "xxx"+m.str[0:len(m.str)-1]; - rune, size = utf8.DecodeRuneInString(s, 3); - if rune != RuneError || size != wantsize { - t.Errorf("DecodeRuneInString(%q, 3) = 0x%04x, %d want 0x%04x, %d", s, rune, size, RuneError, wantsize); - } - - // make sure bad sequences fail - if len(b) == 1 { - b[0] = 0x80; - } else { - b[len(b)-1] = 0x7F; - } - rune, size = utf8.DecodeRune(b); - if rune != RuneError || size != 1 { - t.Errorf("DecodeRune(%q) = 0x%04x, %d want 0x%04x, %d", b, rune, size, RuneError, 1); - } - s = "xxxx"+string(b); - rune, size = utf8.DecodeRune(b); - if rune != RuneError || size != 1 { - t.Errorf("DecodeRuneInString(%q, 4) = 0x%04x, %d want 0x%04x, %d", s, rune, size, RuneError, 1); - } - } -} - -type RuneCountTest struct { - in string; - out int; -} -var runecounttests = []RuneCountTest { - RuneCountTest{ "abcd", 4 }, - RuneCountTest{ "☺☻☹", 3 }, - RuneCountTest{ "1,2,3,4", 7 }, - RuneCountTest{ "\xe2\x00", 2 }, -} -func TestRuneCount(t *testing.T) { - for i := 0; i < len(runecounttests); i++ { - tt := runecounttests[i]; - if out := utf8.RuneCountInString(tt.in); out != tt.out { - t.Errorf("RuneCountInString(%q) = %d, want %d", tt.in, out, tt.out); - } - if out := utf8.RuneCount(bytes(tt.in)); out != tt.out { - t.Errorf("RuneCount(%q) = %d, want %d", tt.in, out, tt.out); - } - } -} |