* Copyright (C) 1999,2000,2001 by Lineo, inc. and Mark Whitley
* Copyright (C) 1999,2000,2001 by Mark Whitley <markw@codepoet.org>
* Copyright (C) 2002 Matt Kraai
- * Copyright (C) 2003 by Glenn McGrath <bug1@iinet.net.au>
+ * Copyright (C) 2003 by Glenn McGrath
* Copyright (C) 2003,2004 by Rob Landley <rob@landley.net>
*
* MAINTAINER: Rob Landley <rob@landley.net>
add_cmd() is called on each line of sed command text (from a file or from
the command line). It calls get_address() and parse_cmd_args(). The
resulting sed_cmd_t structures are appended to a linked list
- (bbg.sed_cmd_head/bbg.sed_cmd_tail).
+ (G.sed_cmd_head/G.sed_cmd_tail).
add_input_file() adds a FILE * to the list of input files. We need to
know all input sources ahead of time to find the last line for the $ match.
Reference http://www.opengroup.org/onlinepubs/007904975/utilities/sed.html
*/
-#include "busybox.h"
+#include "libbb.h"
#include "xregex.h"
/* Each sed command turns into one of these structures. */
typedef struct sed_cmd_s {
/* Ordered by alignment requirements: currently 36 bytes on x86 */
+ struct sed_cmd_s *next; /* Next command (linked list, NULL terminated) */
/* address storage */
regex_t *beg_match; /* sed -e '/match/cmd' */
int beg_line; /* 'sed 1p' 0 == apply commands to all lines */
int end_line; /* 'sed 1,3p' 0 == one line only. -1 = last line ($) */
- FILE *file; /* File (sw) command writes to, -1 for none. */
+ FILE *sw_file; /* File (sw) command writes to, -1 for none. */
char *string; /* Data string for (saicytb) commands. */
- unsigned short which_match; /* (s) Which match to replace (0 for all) */
+ unsigned which_match; /* (s) Which match to replace (0 for all) */
/* Bitfields (gcc won't group them if we don't) */
- unsigned int invert:1; /* the '!' after the address */
- unsigned int in_match:1; /* Next line also included in match? */
- unsigned int sub_p:1; /* (s) print option */
+ unsigned invert:1; /* the '!' after the address */
+ unsigned in_match:1; /* Next line also included in match? */
+ unsigned sub_p:1; /* (s) print option */
- int last_char; /* Last line written by (sw) had no '\n' */
+ char sw_last_char; /* Last line written by (sw) had no '\n' */
/* GENERAL FIELDS */
char cmd; /* The command char: abcdDgGhHilnNpPqrstwxy:={} */
- struct sed_cmd_s *next; /* Next command (linked list, NULL terminated) */
} sed_cmd_t;
-static const char *const semicolon_whitespace = "; \n\r\t\v";
+static const char semicolon_whitespace[] ALIGN1 = "; \n\r\t\v";
-struct sed_globals {
+struct globals {
/* options */
int be_quiet, regex_type;
FILE *nonstdout;
int idx; /* Space used */
int len; /* Space allocated */
} pipeline;
-} bbg;
+};
+#define G (*(struct globals*)&bb_common_bufsiz1)
+void BUG_sed_globals_too_big(void);
+#define INIT_G() do { \
+ if (sizeof(struct globals) > COMMON_BUFSIZE) \
+ BUG_sed_globals_too_big(); \
+ G.sed_cmd_tail = &G.sed_cmd_head; \
+} while (0)
#if ENABLE_FEATURE_CLEAN_UP
static void sed_free_and_close_stuff(void)
{
- sed_cmd_t *sed_cmd = bbg.sed_cmd_head.next;
+ sed_cmd_t *sed_cmd = G.sed_cmd_head.next;
- llist_free(bbg.append_head, free);
+ llist_free(G.append_head, free);
while (sed_cmd) {
sed_cmd_t *sed_cmd_next = sed_cmd->next;
- if (sed_cmd->file)
- xprint_and_close_file(sed_cmd->file);
+ if (sed_cmd->sw_file)
+ xprint_and_close_file(sed_cmd->sw_file);
if (sed_cmd->beg_match) {
regfree(sed_cmd->beg_match);
sed_cmd = sed_cmd_next;
}
- if (bbg.hold_space) free(bbg.hold_space);
+ free(G.hold_space);
- while (bbg.current_input_file < bbg.input_file_count)
- fclose(bbg.input_file_list[bbg.current_input_file++]);
+ while (G.current_input_file < G.input_file_count)
+ fclose(G.input_file_list[G.current_input_file++]);
}
#else
void sed_free_and_close_stuff(void);
static void cleanup_outname(void)
{
- if (bbg.outname) unlink(bbg.outname);
+ if (G.outname) unlink(G.outname);
}
-/* strdup, replacing "\n" with '\n', and "\delimiter" with 'delimiter' */
+/* strcpy, replacing "\from" with 'to'. If to is NUL, replacing "\any" with 'any' */
-static void parse_escapes(char *dest, char *string, int len, char from, char to)
+static void parse_escapes(char *dest, const char *string, int len, char from, char to)
{
int i = 0;
while (i < len) {
if (string[i] == '\\') {
if (!to || string[i+1] == from) {
- *(dest++) = to ? to : string[i+1];
+ *dest++ = to ? to : string[i+1];
i += 2;
continue;
- } else *(dest++) = string[i++];
+ }
+ *dest++ = string[i++];
}
- *(dest++) = string[i++];
+ /* TODO: is it safe wrt a string with trailing '\\' ? */
+ *dest++ = string[i++];
}
- *dest = 0;
+ *dest = '\0';
}
-static char *copy_parsing_escapes(char *string, int len)
+static char *copy_parsing_escapes(const char *string, int len)
{
char *dest = xmalloc(len + 1);
parse_escapes(dest, string, len, 'n', '\n');
+ /* GNU sed also recognizes \t */
+ parse_escapes(dest, dest, strlen(dest), 't', '\t');
return dest;
}
/*
* index_of_next_unescaped_regexp_delim - walks left to right through a string
* beginning at a specified index and returns the index of the next regular
- * expression delimiter (typically a forward * slash ('/')) not preceded by
+ * expression delimiter (typically a forward slash ('/')) not preceded by
* a backslash ('\'). A negative delimiter disables square bracket checking.
*/
-static int index_of_next_unescaped_regexp_delim(int delimiter, char *str)
+static int index_of_next_unescaped_regexp_delim(int delimiter, const char *str)
{
int bracket = -1;
int escaped = 0;
/*
* Returns the index of the third delimiter
*/
-static int parse_regex_delim(char *cmdstr, char **match, char **replace)
+static int parse_regex_delim(const char *cmdstr, char **match, char **replace)
{
- char *cmdstr_ptr = cmdstr;
+ const char *cmdstr_ptr = cmdstr;
char delimiter;
int idx = 0;
/*
* returns the index in the string just past where the address ends.
*/
-static int get_address(char *my_str, int *linenum, regex_t ** regex)
+static int get_address(const char *my_str, int *linenum, regex_t ** regex)
{
- char *pos = my_str;
+ const char *pos = my_str;
if (isdigit(*my_str)) {
- *linenum = strtol(my_str, &pos, 10);
+ *linenum = strtol(my_str, (char**)&pos, 10);
/* endstr shouldnt ever equal NULL */
} else if (*my_str == '$') {
*linenum = -1;
next = index_of_next_unescaped_regexp_delim(delimiter, ++pos);
temp = copy_parsing_escapes(pos, next);
*regex = xmalloc(sizeof(regex_t));
- xregcomp(*regex, temp, bbg.regex_type|REG_NEWLINE);
+ xregcomp(*regex, temp, G.regex_type|REG_NEWLINE);
free(temp);
/* Move position to next character after last delimiter */
pos += (next+1);
}
/* Grab a filename. Whitespace at start is skipped, then goes to EOL. */
-static int parse_file_cmd(sed_cmd_t *sed_cmd, char *filecmdstr, char **retval)
+static int parse_file_cmd(sed_cmd_t *sed_cmd, const char *filecmdstr, char **retval)
{
int start = 0, idx, hack = 0;
/* Skip whitespace, then grab filename to end of line */
- while (isspace(filecmdstr[start])) start++;
+ while (isspace(filecmdstr[start]))
+ start++;
idx = start;
- while (filecmdstr[idx] && filecmdstr[idx] != '\n') idx++;
+ while (filecmdstr[idx] && filecmdstr[idx] != '\n')
+ idx++;
/* If lines glued together, put backslash back. */
- if (filecmdstr[idx] == '\n') hack = 1;
+ if (filecmdstr[idx] == '\n')
+ hack = 1;
if (idx == start)
bb_error_msg_and_die("empty filename");
*retval = xstrndup(filecmdstr+start, idx-start+hack+1);
- if (hack) (*retval)[idx] = '\\';
+ if (hack)
+ (*retval)[idx] = '\\';
return idx;
}
-static int parse_subst_cmd(sed_cmd_t *sed_cmd, char *substr)
+static int parse_subst_cmd(sed_cmd_t *sed_cmd, const char *substr)
{
- int cflags = bbg.regex_type;
+ int cflags = G.regex_type;
char *match;
- int idx = 0;
+ int idx;
/*
* A substitution command should look something like this:
if (isdigit(substr[idx])) {
if (match[0] != '^') {
/* Match 0 treated as all, multiple matches we take the last one. */
- char *pos = substr + idx;
- /* FIXME: error check? */
- sed_cmd->which_match = (unsigned short)strtol(substr+idx, &pos, 10);
+ const char *pos = substr + idx;
+/* FIXME: error check? */
+ sed_cmd->which_match = (unsigned)strtol(substr+idx, (char**) &pos, 10);
idx = pos - substr;
}
continue;
switch (substr[idx]) {
/* Replace all occurrences */
case 'g':
- if (match[0] != '^') sed_cmd->which_match = 0;
+ if (match[0] != '^')
+ sed_cmd->which_match = 0;
break;
/* Print pattern space */
case 'p':
{
char *temp;
idx += parse_file_cmd(sed_cmd, substr+idx, &temp);
-
break;
}
/* Ignore case (gnu exension) */
/*
* Process the commands arguments
*/
-static char *parse_cmd_args(sed_cmd_t *sed_cmd, char *cmdstr)
+static const char *parse_cmd_args(sed_cmd_t *sed_cmd, const char *cmdstr)
{
/* handle (s)ubstitution command */
if (sed_cmd->cmd == 's')
break;
}
sed_cmd->string = xstrdup(cmdstr);
- parse_escapes(sed_cmd->string, sed_cmd->string, strlen(cmdstr), 0, 0);
+ /* "\anychar" -> "anychar" */
+ parse_escapes(sed_cmd->string, sed_cmd->string, strlen(cmdstr), '\0', '\0');
cmdstr += strlen(cmdstr);
/* handle file cmds: (r)ead */
} else if (strchr("rw", sed_cmd->cmd)) {
if (sed_cmd->end_line || sed_cmd->end_match)
bb_error_msg_and_die("command only uses one address");
cmdstr += parse_file_cmd(sed_cmd, cmdstr, &sed_cmd->string);
- if (sed_cmd->cmd == 'w')
- sed_cmd->file = xfopen(sed_cmd->string, "w");
+ if (sed_cmd->cmd == 'w') {
+ sed_cmd->sw_file = xfopen(sed_cmd->string, "w");
+ sed_cmd->sw_last_char = '\n';
+ }
/* handle branch commands */
} else if (strchr(":btT", sed_cmd->cmd)) {
int length;
/* Parse address+command sets, skipping comment lines. */
-static void add_cmd(char *cmdstr)
+static void add_cmd(const char *cmdstr)
{
sed_cmd_t *sed_cmd;
int temp;
/* Append this line to any unfinished line from last time. */
- if (bbg.add_cmd_line) {
- cmdstr = xasprintf("%s\n%s", bbg.add_cmd_line, cmdstr);
- free(bbg.add_cmd_line);
- bbg.add_cmd_line = cmdstr;
+ if (G.add_cmd_line) {
+ char *tp = xasprintf("%s\n%s", G.add_cmd_line, cmdstr);
+ free(G.add_cmd_line);
+ cmdstr = G.add_cmd_line = tp;
}
/* If this line ends with backslash, request next line. */
temp = strlen(cmdstr);
- if (temp && cmdstr[temp-1] == '\\') {
- if (!bbg.add_cmd_line)
- bbg.add_cmd_line = xstrdup(cmdstr);
- bbg.add_cmd_line[temp-1] = 0;
+ if (temp && cmdstr[--temp] == '\\') {
+ if (!G.add_cmd_line)
+ G.add_cmd_line = xstrdup(cmdstr);
+ G.add_cmd_line[temp] = '\0';
return;
}
if (*cmdstr == '#') {
/* "#n" is the same as using -n on the command line */
if (cmdstr[1] == 'n')
- bbg.be_quiet++;
+ G.be_quiet++;
cmdstr = strpbrk(cmdstr, "\n\r");
if (!cmdstr) break;
continue;
cmdstr = parse_cmd_args(sed_cmd, cmdstr);
/* Add the command to the command array */
- bbg.sed_cmd_tail->next = sed_cmd;
- bbg.sed_cmd_tail = bbg.sed_cmd_tail->next;
+ G.sed_cmd_tail->next = sed_cmd;
+ G.sed_cmd_tail = G.sed_cmd_tail->next;
}
/* If we glued multiple lines together, free the memory. */
- free(bbg.add_cmd_line);
- bbg.add_cmd_line = NULL;
+ free(G.add_cmd_line);
+ G.add_cmd_line = NULL;
}
/* Append to a string, reallocating memory as necessary. */
static void pipe_putc(char c)
{
- if (bbg.pipeline.idx == bbg.pipeline.len) {
- bbg.pipeline.buf = xrealloc(bbg.pipeline.buf,
- bbg.pipeline.len + PIPE_GROW);
- bbg.pipeline.len += PIPE_GROW;
+ if (G.pipeline.idx == G.pipeline.len) {
+ G.pipeline.buf = xrealloc(G.pipeline.buf,
+ G.pipeline.len + PIPE_GROW);
+ G.pipeline.len += PIPE_GROW;
}
- bbg.pipeline.buf[bbg.pipeline.idx++] = c;
+ G.pipeline.buf[G.pipeline.idx++] = c;
}
static void do_subst_w_backrefs(char *line, char *replace)
/* go through the replacement string */
for (i = 0; replace[i]; i++) {
/* if we find a backreference (\1, \2, etc.) print the backref'ed * text */
- if (replace[i] == '\\' && replace[i+1] >= '0' && replace[i+1] <= '9') {
- int backref = replace[++i]-'0';
-
- /* print out the text held in bbg.regmatch[backref] */
- if (bbg.regmatch[backref].rm_so != -1) {
- j = bbg.regmatch[backref].rm_so;
- while (j < bbg.regmatch[backref].rm_eo)
- pipe_putc(line[j++]);
+ if (replace[i] == '\\') {
+ unsigned backref = replace[++i] - '0';
+ if (backref <= 9) {
+ /* print out the text held in G.regmatch[backref] */
+ if (G.regmatch[backref].rm_so != -1) {
+ j = G.regmatch[backref].rm_so;
+ while (j < G.regmatch[backref].rm_eo)
+ pipe_putc(line[j++]);
+ }
+ continue;
}
+ /* I _think_ it is impossible to get '\' to be
+ * the last char in replace string. Thus we dont check
+ * for replace[i] == NUL. (counterexample anyone?) */
+ /* if we find a backslash escaped character, print the character */
+ pipe_putc(replace[i]);
+ continue;
}
-
- /* if we find a backslash escaped character, print the character */
- else if (replace[i] == '\\') pipe_putc(replace[++i]);
-
/* if we find an unescaped '&' print out the whole matched text. */
- else if (replace[i] == '&') {
- j = bbg.regmatch[0].rm_so;
- while (j < bbg.regmatch[0].rm_eo)
+ if (replace[i] == '&') {
+ j = G.regmatch[0].rm_so;
+ while (j < G.regmatch[0].rm_eo)
pipe_putc(line[j++]);
+ continue;
}
/* Otherwise just output the character. */
- else pipe_putc(replace[i]);
+ pipe_putc(replace[i]);
}
}
/* Handle empty regex. */
if (sed_cmd->sub_match == NULL) {
- current_regex = bbg.previous_regex_ptr;
+ current_regex = G.previous_regex_ptr;
if (!current_regex)
bb_error_msg_and_die("no previous regexp");
} else
- bbg.previous_regex_ptr = current_regex = sed_cmd->sub_match;
+ G.previous_regex_ptr = current_regex = sed_cmd->sub_match;
/* Find the first match */
- if (REG_NOMATCH == regexec(current_regex, oldline, 10, bbg.regmatch, 0))
+ if (REG_NOMATCH == regexec(current_regex, oldline, 10, G.regmatch, 0))
return 0;
/* Initialize temporary output buffer. */
- bbg.pipeline.buf = xmalloc(PIPE_GROW);
- bbg.pipeline.len = PIPE_GROW;
- bbg.pipeline.idx = 0;
+ G.pipeline.buf = xmalloc(PIPE_GROW);
+ G.pipeline.len = PIPE_GROW;
+ G.pipeline.idx = 0;
/* Now loop through, substituting for matches */
do {
echo " a.b" | busybox sed 's [^ .]* x g'
The match_count check is so not to break
echo "hi" | busybox sed 's/^/!/g' */
- if (!bbg.regmatch[0].rm_so && !bbg.regmatch[0].rm_eo && match_count) {
+ if (!G.regmatch[0].rm_so && !G.regmatch[0].rm_eo && match_count) {
pipe_putc(*oldline++);
continue;
}
/* If we aren't interested in this match, output old line to
end of match and continue */
if (sed_cmd->which_match && sed_cmd->which_match != match_count) {
- for (i = 0; i < bbg.regmatch[0].rm_eo; i++)
+ for (i = 0; i < G.regmatch[0].rm_eo; i++)
pipe_putc(*oldline++);
continue;
}
/* print everything before the match */
- for (i = 0; i < bbg.regmatch[0].rm_so; i++)
+ for (i = 0; i < G.regmatch[0].rm_so; i++)
pipe_putc(oldline[i]);
/* then print the substitution string */
do_subst_w_backrefs(oldline, sed_cmd->string);
/* advance past the match */
- oldline += bbg.regmatch[0].rm_eo;
+ oldline += G.regmatch[0].rm_eo;
/* flag that something has changed */
altered++;
/* if we're not doing this globally, get out now */
- if (sed_cmd->which_match) break;
- } while (*oldline && (regexec(current_regex, oldline, 10, bbg.regmatch, 0) != REG_NOMATCH));
+ if (sed_cmd->which_match)
+ break;
+ } while (*oldline && (regexec(current_regex, oldline, 10, G.regmatch, 0) != REG_NOMATCH));
/* Copy rest of string into output pipeline */
pipe_putc(0);
free(*line);
- *line = bbg.pipeline.buf;
+ *line = G.pipeline.buf;
return altered;
}
{
sed_cmd_t *sed_cmd;
- for (sed_cmd = bbg.sed_cmd_head.next; sed_cmd; sed_cmd = sed_cmd->next) {
+ for (sed_cmd = G.sed_cmd_head.next; sed_cmd; sed_cmd = sed_cmd->next) {
if (sed_cmd->cmd == ':' && sed_cmd->string && !strcmp(sed_cmd->string, label)) {
return sed_cmd;
}
static void append(char *s)
{
- llist_add_to_end(&bbg.append_head, xstrdup(s));
+ llist_add_to_end(&G.append_head, xstrdup(s));
}
static void flush_append(void)
char *data;
/* Output appended lines. */
- while ((data = (char *)llist_pop(&bbg.append_head))) {
- fprintf(bbg.nonstdout, "%s\n", data);
+ while ((data = (char *)llist_pop(&G.append_head))) {
+ fprintf(G.nonstdout, "%s\n", data);
free(data);
}
}
static void add_input_file(FILE *file)
{
- bbg.input_file_list = xrealloc(bbg.input_file_list,
- (bbg.input_file_count + 1) * sizeof(FILE *));
- bbg.input_file_list[bbg.input_file_count++] = file;
+ G.input_file_list = xrealloc(G.input_file_list,
+ (G.input_file_count + 1) * sizeof(FILE *));
+ G.input_file_list[G.input_file_count++] = file;
}
-/* Get next line of input from bbg.input_file_list, flushing append buffer and
+/* Get next line of input from G.input_file_list, flushing append buffer and
* noting if we ran out of files without a newline on the last line we read.
*/
-static char *get_next_line(int *last_char)
+enum {
+ NO_EOL_CHAR = 1,
+ LAST_IS_NUL = 2,
+};
+static char *get_next_line(char *gets_char)
{
char *temp = NULL;
- int len, lc;
+ int len;
+ char gc;
- lc = 0;
flush_append();
- while (bbg.current_input_file < bbg.input_file_count) {
- temp = bb_get_chunk_from_file(
- bbg.input_file_list[bbg.current_input_file], &len);
+
+ /* will be returned if last line in the file
+ * doesn't end with either '\n' or '\0' */
+ gc = NO_EOL_CHAR;
+ while (G.current_input_file < G.input_file_count) {
+ FILE *fp = G.input_file_list[G.current_input_file];
+ /* Read line up to a newline or NUL byte, inclusive,
+ * return malloc'ed char[]. length of the chunk read
+ * is stored in len. NULL if EOF/error */
+ temp = bb_get_chunk_from_file(fp, &len);
if (temp) {
/* len > 0 here, it's ok to do temp[len-1] */
char c = temp[len-1];
if (c == '\n' || c == '\0') {
temp[len-1] = '\0';
- lc |= (unsigned char)c;
- break;
+ gc = c;
+ if (c == '\0') {
+ int ch = fgetc(fp);
+ if (ch != EOF)
+ ungetc(ch, fp);
+ else
+ gc = LAST_IS_NUL;
+ }
}
- /* will be returned if last line in the file
- * doesn't end with either '\n' or '\0' */
- lc |= 0x100;
+ /* else we put NO_EOL_CHAR into *gets_char */
break;
+
+ /* NB: I had the idea of peeking next file(s) and returning
+ * NO_EOL_CHAR only if it is the *last* non-empty
+ * input file. But there is a case where this won't work:
+ * file1: "a woo\nb woo"
+ * file2: "c no\nd no"
+ * sed -ne 's/woo/bang/p' input1 input2 => "a bang\nb bang"
+ * (note: *no* newline after "b bang"!) */
}
/* Close this file and advance to next one */
- fclose(bbg.input_file_list[bbg.current_input_file++]);
- /* "this is the first line from new input file" */
- lc |= 0x200;
+ fclose(fp);
+ G.current_input_file++;
}
- *last_char = lc;
+ *gets_char = gc;
return temp;
}
/* Output line of text. */
/* Note:
- * The tricks with 0x200 and last_puts_char are there to emulate gnu sed.
+ * The tricks with NO_EOL_CHAR and last_puts_char are there to emulate gnu sed.
* Without them, we had this:
* echo -n thingy >z1
* echo -n again >z2
* >znull
- * sed "s/i/z/" z1 z2 znull | hexdump -vC output:
+ * sed "s/i/z/" z1 z2 znull | hexdump -vC
+ * output:
* gnu sed 4.1.5:
* 00000000 74 68 7a 6e 67 79 0a 61 67 61 7a 6e |thzngy.agazn|
* bbox:
* 00000000 74 68 7a 6e 67 79 61 67 61 7a 6e |thzngyagazn|
*/
-
-static int puts_maybe_newline(char *s, FILE *file, int prev_last_char, int last_char)
+static void puts_maybe_newline(char *s, FILE *file, char *last_puts_char, char last_gets_char)
{
- static char last_puts_char;
+ char lpc = *last_puts_char;
- /* Is this a first line from new file
- * and old file didn't end with '\n'? */
- if ((last_char & 0x200) && last_puts_char != '\n') {
+ /* Need to insert a '\n' between two files because first file's
+ * last line wasn't terminated? */
+ if (lpc != '\n' && lpc != '\0') {
fputc('\n', file);
- last_puts_char = '\n';
+ lpc = '\n';
}
fputs(s, file);
- /* 'x': we don't care what is it, but we know it isn't '\n' */
- if (s[0]) last_puts_char = 'x';
- if (!(last_char & 0x100)) { /* had trailing '\n' or '\0'? */
- last_char &= 0xff;
- fputc(last_char, file);
- last_puts_char = last_char;
+
+ /* 'x' - just something which is not '\n', '\0' or NO_EOL_CHAR */
+ if (s[0])
+ lpc = 'x';
+
+ /* had trailing '\0' and it was last char of file? */
+ if (last_gets_char == LAST_IS_NUL) {
+ fputc('\0', file);
+ lpc = 'x'; /* */
+ } else
+ /* had trailing '\n' or '\0'? */
+ if (last_gets_char != NO_EOL_CHAR) {
+ fputc(last_gets_char, file);
+ lpc = last_gets_char;
}
if (ferror(file)) {
xfunc_error_retval = 4; /* It's what gnu sed exits with... */
bb_error_msg_and_die(bb_msg_write_error);
}
-
- return last_char;
+ *last_puts_char = lpc;
}
-#define sed_puts(s, n) \
- (prev_last_char = puts_maybe_newline(s, bbg.nonstdout, prev_last_char, n))
+#define sed_puts(s, n) (puts_maybe_newline(s, G.nonstdout, &last_puts_char, n))
+
+static int beg_match(sed_cmd_t *sed_cmd, const char *pattern_space)
+{
+ int retval = sed_cmd->beg_match && !regexec(sed_cmd->beg_match, pattern_space, 0, NULL, 0);
+ if (retval)
+ G.previous_regex_ptr = sed_cmd->beg_match;
+ return retval;
+}
/* Process all the lines in all the files */
static void process_files(void)
{
char *pattern_space, *next_line;
- int linenum = 0, prev_last_char = 0;
- int last_char, next_last_char = 0;
+ int linenum = 0;
+ char last_puts_char = '\n';
+ char last_gets_char, next_gets_char;
sed_cmd_t *sed_cmd;
int substituted;
/* Prime the pump */
- next_line = get_next_line(&next_last_char);
+ next_line = get_next_line(&next_gets_char);
/* go through every line in each file */
again:
/* Advance to next line. Stop if out of lines. */
pattern_space = next_line;
if (!pattern_space) return;
- last_char = next_last_char;
+ last_gets_char = next_gets_char;
/* Read one line in advance so we can act on the last line,
* the '$' address */
- next_line = get_next_line(&next_last_char);
+ next_line = get_next_line(&next_gets_char);
linenum++;
restart:
/* for every line, go through all the commands */
- for (sed_cmd = bbg.sed_cmd_head.next; sed_cmd; sed_cmd = sed_cmd->next) {
+ for (sed_cmd = G.sed_cmd_head.next; sed_cmd; sed_cmd = sed_cmd->next) {
int old_matched, matched;
old_matched = sed_cmd->in_match;
/* Or did we match the start of a numerical range? */
|| (sed_cmd->beg_line > 0 && (sed_cmd->beg_line == linenum))
/* Or does this line match our begin address regex? */
- || (sed_cmd->beg_match &&
- !regexec(sed_cmd->beg_match, pattern_space, 0, NULL, 0))
+ || (beg_match(sed_cmd, pattern_space))
/* Or did we match last line of input? */
|| (sed_cmd->beg_line == -1 && next_line == NULL);
/* or does this line matches our last address regex */
|| (sed_cmd->end_match && old_matched
&& (regexec(sed_cmd->end_match,
- pattern_space, 0, NULL, 0) == 0))
+ pattern_space, 0, NULL, 0) == 0))
);
}
/* Skip blocks of commands we didn't match. */
if (sed_cmd->cmd == '{') {
- if (sed_cmd->invert ? matched : !matched)
- while (sed_cmd && sed_cmd->cmd != '}')
+ if (sed_cmd->invert ? matched : !matched) {
+ while (sed_cmd->cmd != '}') {
sed_cmd = sed_cmd->next;
- if (!sed_cmd) bb_error_msg_and_die("unterminated {");
+ if (!sed_cmd)
+ bb_error_msg_and_die("unterminated {");
+ }
+ }
continue;
}
if (sed_cmd->invert ? !matched : matched) {
/* Update last used regex in case a blank substitute BRE is found */
if (sed_cmd->beg_match) {
- bbg.previous_regex_ptr = sed_cmd->beg_match;
+ G.previous_regex_ptr = sed_cmd->beg_match;
}
/* actual sedding */
/* Print line number */
case '=':
- fprintf(bbg.nonstdout, "%d\n", linenum);
+ fprintf(G.nonstdout, "%d\n", linenum);
break;
/* Write the current pattern space up to the first newline */
if (tmp) {
*tmp = '\0';
- sed_puts(pattern_space, 1);
+ /* TODO: explain why '\n' below */
+ sed_puts(pattern_space, '\n');
*tmp = '\n';
break;
}
/* Write the current pattern space to output */
case 'p':
- sed_puts(pattern_space, last_char);
+ /* NB: we print this _before_ the last line
+ * (of current file) is printed. Even if
+ * that line is nonterminated, we print
+ * '\n' here (gnu sed does the same) */
+ sed_puts(pattern_space, '\n');
break;
/* Delete up through first newline */
case 'D':
/* handle p option */
if (sed_cmd->sub_p)
- sed_puts(pattern_space, last_char);
+ sed_puts(pattern_space, last_gets_char);
/* handle w option */
- if (sed_cmd->file)
- sed_cmd->last_char = puts_maybe_newline(
- pattern_space, sed_cmd->file,
- sed_cmd->last_char, last_char);
+ if (sed_cmd->sw_file)
+ puts_maybe_newline(
+ pattern_space, sed_cmd->sw_file,
+ &sed_cmd->sw_last_char, last_gets_char);
break;
/* Append line to linked list to be printed later */
/* Insert text before this line */
case 'i':
- sed_puts(sed_cmd->string, 1);
+ sed_puts(sed_cmd->string, '\n');
break;
/* Cut and paste text (replace) */
case 'c':
/* Only triggers on last line of a matching range. */
if (!sed_cmd->in_match)
- sed_puts(sed_cmd->string, 0);
+ sed_puts(sed_cmd->string, NO_EOL_CHAR);
goto discard_line;
/* Read file, append contents to output */
/* Write pattern space to file. */
case 'w':
- sed_cmd->last_char = puts_maybe_newline(
- pattern_space, sed_cmd->file,
- sed_cmd->last_char, last_char);
+ puts_maybe_newline(
+ pattern_space, sed_cmd->sw_file,
+ &sed_cmd->sw_last_char, last_gets_char);
break;
/* Read next line from input */
case 'n':
- if (!bbg.be_quiet)
- sed_puts(pattern_space, last_char);
+ if (!G.be_quiet)
+ sed_puts(pattern_space, last_gets_char);
if (next_line) {
free(pattern_space);
pattern_space = next_line;
- last_char = next_last_char;
- next_line = get_next_line(&next_last_char);
+ last_gets_char = next_gets_char;
+ next_line = get_next_line(&next_gets_char);
linenum++;
break;
}
pattern_space = realloc(pattern_space, len + strlen(next_line) + 2);
pattern_space[len] = '\n';
strcpy(pattern_space + len+1, next_line);
- last_char = next_last_char;
- next_line = get_next_line(&next_last_char);
+ last_gets_char = next_gets_char;
+ next_line = get_next_line(&next_gets_char);
linenum++;
break;
}
}
case 'g': /* Replace pattern space with hold space */
free(pattern_space);
- pattern_space = xstrdup(bbg.hold_space ? bbg.hold_space : "");
+ pattern_space = xstrdup(G.hold_space ? G.hold_space : "");
break;
case 'G': /* Append newline and hold space to pattern space */
{
if (pattern_space)
pattern_space_size += strlen(pattern_space);
- if (bbg.hold_space)
- hold_space_size = strlen(bbg.hold_space);
+ if (G.hold_space)
+ hold_space_size = strlen(G.hold_space);
pattern_space = xrealloc(pattern_space,
pattern_space_size + hold_space_size);
if (pattern_space_size == 2)
pattern_space[0] = 0;
strcat(pattern_space, "\n");
- if (bbg.hold_space)
- strcat(pattern_space, bbg.hold_space);
- last_char = '\n';
+ if (G.hold_space)
+ strcat(pattern_space, G.hold_space);
+ last_gets_char = '\n';
break;
}
case 'h': /* Replace hold space with pattern space */
- free(bbg.hold_space);
- bbg.hold_space = xstrdup(pattern_space);
+ free(G.hold_space);
+ G.hold_space = xstrdup(pattern_space);
break;
case 'H': /* Append newline and pattern space to hold space */
{
int hold_space_size = 2;
int pattern_space_size = 0;
- if (bbg.hold_space)
- hold_space_size += strlen(bbg.hold_space);
+ if (G.hold_space)
+ hold_space_size += strlen(G.hold_space);
if (pattern_space)
pattern_space_size = strlen(pattern_space);
- bbg.hold_space = xrealloc(bbg.hold_space,
+ G.hold_space = xrealloc(G.hold_space,
hold_space_size + pattern_space_size);
if (hold_space_size == 2)
- *bbg.hold_space = 0;
- strcat(bbg.hold_space, "\n");
+ *G.hold_space = 0;
+ strcat(G.hold_space, "\n");
if (pattern_space)
- strcat(bbg.hold_space, pattern_space);
+ strcat(G.hold_space, pattern_space);
break;
}
case 'x': /* Exchange hold and pattern space */
{
char *tmp = pattern_space;
- pattern_space = bbg.hold_space ? : xzalloc(1);
- last_char = '\n';
- bbg.hold_space = tmp;
+ pattern_space = G.hold_space ? : xzalloc(1);
+ last_gets_char = '\n';
+ G.hold_space = tmp;
break;
}
}
/*
* exit point from sedding...
*/
-discard_commands:
+ discard_commands:
/* we will print the line unless we were told to be quiet ('-n')
or if the line was suppressed (ala 'd'elete) */
- if (!bbg.be_quiet) sed_puts(pattern_space, last_char);
+ if (!G.be_quiet)
+ sed_puts(pattern_space, last_gets_char);
/* Delete and such jump here. */
-discard_line:
+ discard_line:
flush_append();
free(pattern_space);
}
/* It is possible to have a command line argument with embedded
- newlines. This counts as multiple command lines. */
+ * newlines. This counts as multiple command lines.
+ * However, newline can be escaped: 's/e/z\<newline>z/'
+ * We check for this.
+ */
static void add_cmd_block(char *cmdstr)
{
- int go = 1;
- char *temp = xstrdup(cmdstr), *temp2 = temp;
-
- while (go) {
- int len = strcspn(temp2, "\n");
- if (!temp2[len]) go = 0;
- else temp2[len] = 0;
- add_cmd(temp2);
- temp2 += len+1;
- }
- free(temp);
-}
+ char *sv, *eol;
-static void add_cmds_link(llist_t *opt_e)
-{
- if (!opt_e) return;
- add_cmds_link(opt_e->link);
- add_cmd_block(opt_e->data);
- free(opt_e);
-}
-
-static void add_files_link(llist_t *opt_f)
-{
- char *line;
- FILE *cmdfile;
- if (!opt_f) return;
- add_files_link(opt_f->link);
- cmdfile = xfopen(opt_f->data, "r");
- while ((line = xmalloc_getline(cmdfile)) != NULL) {
- add_cmd(line);
- free(line);
- }
- xprint_and_close_file(cmdfile);
- free(opt_f);
+ cmdstr = sv = xstrdup(cmdstr);
+ do {
+ eol = strchr(cmdstr, '\n');
+ next:
+ if (eol) {
+ /* Count preceding slashes */
+ int slashes = 0;
+ char *sl = eol;
+
+ while (sl != cmdstr && *--sl == '\\')
+ slashes++;
+ /* Odd number of preceding slashes - newline is escaped */
+ if (slashes & 1) {
+ strcpy(eol-1, eol);
+ eol = strchr(eol, '\n');
+ goto next;
+ }
+ *eol = '\0';
+ }
+ add_cmd(cmdstr);
+ cmdstr = eol + 1;
+ } while (eol);
+ free(sv);
}
+int sed_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
int sed_main(int argc, char **argv)
{
enum {
llist_t *opt_e, *opt_f;
int status = EXIT_SUCCESS;
- bbg.sed_cmd_tail = &bbg.sed_cmd_head;
+ INIT_G();
/* destroy command strings on exit */
if (ENABLE_FEATURE_CLEAN_UP) atexit(sed_free_and_close_stuff);
opt_e = opt_f = NULL;
opt_complementary = "e::f::" /* can occur multiple times */
"nn"; /* count -n */
- opt = getopt32(argc, argv, "irne:f:", &opt_e, &opt_f,
- &bbg.be_quiet); /* counter for -n */
+ opt = getopt32(argv, "irne:f:", &opt_e, &opt_f,
+ &G.be_quiet); /* counter for -n */
argc -= optind;
argv += optind;
if (opt & OPT_in_place) { // -i
atexit(cleanup_outname);
}
- if (opt & 0x2) bbg.regex_type |= REG_EXTENDED; // -r
- //if (opt & 0x4) bbg.be_quiet++; // -n
- if (opt & 0x8) { // -e
- /* getopt32 reverses order of arguments, handle it */
- add_cmds_link(opt_e);
+ if (opt & 0x2) G.regex_type |= REG_EXTENDED; // -r
+ //if (opt & 0x4) G.be_quiet++; // -n
+ while (opt_e) { // -e
+ add_cmd_block(opt_e->data);
+ opt_e = opt_e->link;
+ /* we leak opt_e here... */
}
- if (opt & 0x10) { // -f
- /* getopt32 reverses order of arguments, handle it */
- add_files_link(opt_f);
+ while (opt_f) { // -f
+ char *line;
+ FILE *cmdfile;
+ cmdfile = xfopen(opt_f->data, "r");
+ while ((line = xmalloc_getline(cmdfile)) != NULL) {
+ add_cmd(line);
+ free(line);
+ }
+ fclose(cmdfile);
+ opt_f = opt_f->link;
+ /* we leak opt_f here... */
}
/* if we didn't get a pattern from -e or -f, use argv[0] */
if (!(opt & 0x18)) {
add_cmd("");
/* By default, we write to stdout */
- bbg.nonstdout = stdout;
+ G.nonstdout = stdout;
/* argv[0..(argc-1)] should be names of file to process. If no
* files were specified or '-' was specified, take input from stdin.
continue;
}
- bbg.outname = xasprintf("%sXXXXXX", argv[i]);
- nonstdoutfd = mkstemp(bbg.outname);
+ G.outname = xasprintf("%sXXXXXX", argv[i]);
+ nonstdoutfd = mkstemp(G.outname);
if (-1 == nonstdoutfd)
- bb_error_msg_and_die("no temp file");
- bbg.nonstdout = fdopen(nonstdoutfd, "w");
+ bb_perror_msg_and_die("cannot create temp file %s", G.outname);
+ G.nonstdout = fdopen(nonstdoutfd, "w");
/* Set permissions of output file */
fchmod(nonstdoutfd, statbuf.st_mode);
add_input_file(file);
process_files();
- fclose(bbg.nonstdout);
+ fclose(G.nonstdout);
- bbg.nonstdout = stdout;
+ G.nonstdout = stdout;
/* unlink(argv[i]); */
- // FIXME: error check / message?
- rename(bbg.outname, argv[i]);
- free(bbg.outname);
- bbg.outname = 0;
+ xrename(G.outname, argv[i]);
+ free(G.outname);
+ G.outname = NULL;
}
- if (bbg.input_file_count > bbg.current_input_file)
+ if (G.input_file_count > G.current_input_file)
process_files();
}
projects / oweals / busybox.git / blobdiff