From 02d8e8247fced0d290e59fca2b003ce6cae1ca06 Mon Sep 17 00:00:00 2001 From: Russ Cox Date: Tue, 7 Dec 2010 23:15:22 -0500 Subject: wc: 4-byte utf-8 capable (Erik Quanstrom) R=rsc http://codereview.appspot.com/3437044 --- src/cmd/wc.c | 75 +++++++++++++++++++++++++++++++++++++++++++++++------------- 1 file changed, 59 insertions(+), 16 deletions(-) (limited to 'src/cmd/wc.c') diff --git a/src/cmd/wc.c b/src/cmd/wc.c index 64a791de..60077595 100644 --- a/src/cmd/wc.c +++ b/src/cmd/wc.c @@ -97,12 +97,13 @@ report(uvlong nline, uvlong nword, uvlong nrune, uvlong nbadr, uvlong nchar, cha * following table. If we're not in statesp or statewd when done, the * file ends with a partial rune. * | character - * state |09,20| 0a |00-7f|80-bf|c0-df|e0-ef|f0-ff - * -------+-----+-----+-----+-----+-----+-----+----- - * statesp|ASP |ASPN |AWDW |AWDWX|AC2W |AC3W |AWDWX - * statewd|ASP |ASPN |AWD |AWDX |AC2 |AC3 |AWDX - * statec2|ASPX |ASPNX|AWDX |AWDR |AC2X |AC3X |AWDX - * statec3|ASPX |ASPNX|AWDX |AC2R |AC2X |AC3X |AWDX + * state |09,20| 0a |00-7f|80-bf|c0-df|e0-ef|f0-f7|f8-ff + * -------+-----+-----+-----+-----+-----+-----+-----+----- + * statesp|ASP |ASPN |AWDW |AWDWX|AC2W |AC3W |AC4W |AWDWX + * statewd|ASP |ASPN |AWD |AWDX |AC2 |AC3 |AC4 |AWDX + * statec2|ASPX |ASPNX|AWDX |AWDR |AC2X |AC3X |AC4X |AWDX + * statec3|ASPX |ASPNX|AWDX |AC2R |AC2X |AC3X |AC4X |AWDX + * statec4|ASPX |ASPNX|AWDX |AC3R |AC2X |AC3X |AC4X |AWDX f4 8f bf bf */ enum{ /* actions */ AC2, /* enter statec2 */ @@ -110,8 +111,12 @@ enum{ /* actions */ AC2W, /* enter statec2, count a word */ AC2X, /* enter statec2, count a bad rune */ AC3, /* enter statec3 */ + AC3R, /* enter statec3, don't count a rune */ AC3W, /* enter statec3, count a word */ AC3X, /* enter statec3, count a bad rune */ + AC4, /* enter statec4 */ + AC4W, /* enter statec4, count a word */ + AC4X, /* enter statec4, count a bad rune */ ASP, /* enter statesp */ ASPN, /* enter statesp, count a newline */ ASPNX, /* enter statesp, count a newline, count a bad rune */ @@ -120,7 +125,7 @@ enum{ /* actions */ AWDR, /* enter statewd, don't count a rune */ AWDW, /* enter statewd, count a word */ AWDWX, /* enter statewd, count a word, count a bad rune */ - AWDX /* enter statewd, count a bad rune */ + AWDX, /* enter statewd, count a bad rune */ }; uchar statesp[256]={ /* looking for the start of a word */ AWDW, AWDW, AWDW, AWDW, AWDW, AWDW, AWDW, AWDW, /* 00-07 */ @@ -153,8 +158,8 @@ AC2W, AC2W, AC2W, AC2W, AC2W, AC2W, AC2W, AC2W, /* d0-d7 */ AC2W, AC2W, AC2W, AC2W, AC2W, AC2W, AC2W, AC2W, /* d8-df */ AC3W, AC3W, AC3W, AC3W, AC3W, AC3W, AC3W, AC3W, /* e0-e7 */ AC3W, AC3W, AC3W, AC3W, AC3W, AC3W, AC3W, AC3W, /* e8-ef */ -AWDWX,AWDWX,AWDWX,AWDWX,AWDWX,AWDWX,AWDWX,AWDWX,/* f0-f7 */ -AWDWX,AWDWX,AWDWX,AWDWX,AWDWX,AWDWX,AWDWX,AWDWX/* f8-ff */ +AC4W, AC4W, AC4W, AC4W, AC4W, AC4W, AC4W, AC4W, /* f0-f7 */ +AWDWX,AWDWX,AWDWX,AWDWX,AWDWX,AWDWX,AWDWX,AWDWX,/* f8-ff */ }; uchar statewd[256]={ /* looking for the next character in a word */ AWD, AWD, AWD, AWD, AWD, AWD, AWD, AWD, /* 00-07 */ @@ -187,8 +192,8 @@ AC2, AC2, AC2, AC2, AC2, AC2, AC2, AC2, /* d0-d7 */ AC2, AC2, AC2, AC2, AC2, AC2, AC2, AC2, /* d8-df */ AC3, AC3, AC3, AC3, AC3, AC3, AC3, AC3, /* e0-e7 */ AC3, AC3, AC3, AC3, AC3, AC3, AC3, AC3, /* e8-ef */ -AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* f0-f7 */ -AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX /* f8-ff */ +AC4, AC4, AC4, AC4, AC4, AC4, AC4, AC4, /* f0-f7 */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* f8-ff */ }; uchar statec2[256]={ /* looking for 10xxxxxx to complete a rune */ AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 00-07 */ @@ -221,8 +226,8 @@ AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, /* d0-d7 */ AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, /* d8-df */ AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, /* e0-e7 */ AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, /* e8-ef */ -AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* f0-f7 */ -AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX /* f8-ff */ +AC4X, AC4X, AC4X, AC4X, AC4X, AC4X, AC4X, AC4X, /* f0-f7 */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* f8-ff */ }; uchar statec3[256]={ /* looking for 10xxxxxx,10xxxxxx to complete a rune */ AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 00-07 */ @@ -255,8 +260,42 @@ AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, /* d0-d7 */ AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, /* d8-df */ AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, /* e0-e7 */ AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, /* e8-ef */ -AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* f0-f7 */ -AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX /* f8-ff */ +AC4X, AC4X, AC4X, AC4X, AC4X, AC4X, AC4X, AC4X, /* f0-f7 */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* f8-ff */ +}; +uchar statec4[256]={ /* looking for 10xxxxxx,10xxxxxx,10xxxxxx to complete a rune */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 00-07 */ +AWDX, ASPX, ASPNX,AWDX, AWDX, AWDX, AWDX, AWDX, /* 08-0f */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 10-17 */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 18-1f */ +ASPX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 20-27 */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 28-2f */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 30-37 */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 38-3f */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 40-47 */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 48-4f */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 50-57 */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 58-5f */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 60-67 */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 68-6f */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 70-77 */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 78-7f */ +AC3R, AC3R, AC3R, AC3R, AC3R, AC3R, AC3R, AC3R, /* 80-87 */ +AC3R, AC3R, AC3R, AC3R, AC3R, AC3R, AC3R, AC3R, /* 88-8f */ +AC3R, AC3R, AC3R, AC3R, AC3R, AC3R, AC3R, AC3R, /* 90-97 */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* 98-9f */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* a0-a7 */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* a8-af */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* b0-b7 */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* b8-bf */ +AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, /* c0-c7 */ +AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, /* c8-cf */ +AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, /* d0-d7 */ +AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, AC2X, /* d8-df */ +AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, /* e0-e7 */ +AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, AC3X, /* e8-ef */ +AC4X, AC4X, AC4X, AC4X, AC4X, AC4X, AC4X, AC4X, /* f0-f7 */ +AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, AWDX, /* f8-ff */ }; void count(int f, char *name) @@ -287,8 +326,12 @@ count(int f, char *name) case AC2W: state=statec2; nword++; break; case AC2X: state=statec2; nbadr++; break; case AC3: state=statec3; break; + case AC3R: state=statec3; --nrune; break; case AC3W: state=statec3; nword++; break; case AC3X: state=statec3; nbadr++; break; + case AC4: state=statec4; break; + case AC4W: state=statec4; nword++; break; + case AC4X: state=statec4; nbadr++; break; case ASP: state=statesp; break; case ASPN: state=statesp; nline++; break; case ASPNX: state=statesp; nline++; nbadr++; break; @@ -304,6 +347,6 @@ count(int f, char *name) if(state!=statesp && state!=statewd) nbadr++; if(n<0) - perror(name); + fprint(2, "%s: %r\n", name); report(nline, nword, nrune, nbadr, nchar, name); } -- cgit v1.2.3