📄 sed.c
字号:
} /* handle edit cmds: (a)ppend, (i)nsert, and (c)hange */ else if (strchr("aic", sed_cmd->cmd)) { if ((sed_cmd->end_line || sed_cmd->end_match) && sed_cmd->cmd != 'c') error_msg_and_die("only a beginning address can be specified for edit commands"); idx += parse_edit_cmd(sed_cmd, &cmdstr[idx]); } /* handle file cmds: (r)ead */ else if (sed_cmd->cmd == 'r') { if (sed_cmd->end_line || sed_cmd->end_match) error_msg_and_die("Command only uses one address"); idx += parse_file_cmd(sed_cmd, &cmdstr[idx]); } else { error_msg_and_die("invalid command"); } /* give back whatever's left over */ return (char *)&cmdstr[idx];}static void add_cmd_str(const char * const cmdstr){ char *mystr = (char *)cmdstr; do { /* trim leading whitespace and semicolons */ move_back(mystr, strspn(mystr, semicolon_whitespace)); /* if we ate the whole thing, that means there was just trailing * whitespace or a final / no-op semicolon. either way, get out */ if (strlen(mystr) == 0) return; /* if this is a comment, jump past it and keep going */ if (mystr[0] == '#') { mystr = strpbrk(mystr, "\n\r"); continue; } /* grow the array */ sed_cmds = xrealloc(sed_cmds, sizeof(struct sed_cmd) * (++ncmds)); /* zero new element */ memset(&sed_cmds[ncmds-1], 0, sizeof(struct sed_cmd)); /* load command string into new array element, get remainder */ mystr = parse_cmd_str(&sed_cmds[ncmds-1], mystr); } while (mystr && strlen(mystr));}static void load_cmd_file(char *filename){ FILE *cmdfile; char *line; char *nextline; cmdfile = xfopen(filename, "r"); while ((line = get_line_from_file(cmdfile)) != NULL) { /* if a line ends with '\' it needs the next line appended to it */ while (line[strlen(line)-2] == '\\' && (nextline = get_line_from_file(cmdfile)) != NULL) { line = xrealloc(line, strlen(line) + strlen(nextline) + 1); strcat(line, nextline); free(nextline); } /* eat trailing newline (if any) --if I don't do this, edit commands * (aic) will print an extra newline */ chomp(line); add_cmd_str(line); free(line); }}struct pipeline { char *buf; int idx; int len;};#define PIPE_MAGIC 0x7f#define PIPE_GROW 64 void pipe_putc(struct pipeline *const pipeline, char c){ if (pipeline->buf[pipeline->idx] == PIPE_MAGIC) { pipeline->buf = xrealloc(pipeline->buf, pipeline->len + PIPE_GROW); memset(pipeline->buf + pipeline->len, 0, PIPE_GROW); pipeline->len += PIPE_GROW; pipeline->buf[pipeline->len - 1] = PIPE_MAGIC; } pipeline->buf[pipeline->idx++] = (c);}#define pipeputc(c) pipe_putc(pipeline, c)#if 0{ if (pipeline[pipeline_idx] == PIPE_MAGIC) { \ pipeline = xrealloc(pipeline, pipeline_len+PIPE_GROW); \ memset(pipeline+pipeline_len, 0, PIPE_GROW); \ pipeline_len += PIPE_GROW; \ pipeline[pipeline_len-1] = PIPE_MAGIC; } \ pipeline[pipeline_idx++] = (c); }#endifstatic void print_subst_w_backrefs(const char *line, const char *replace, regmatch_t *regmatch, struct pipeline *const pipeline, int matches){ int i; /* 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] == '\\' && isdigit(replace[i+1])) { int j; char tmpstr[2]; int backref; ++i; /* i now indexes the backref number, instead of the leading slash */ tmpstr[0] = replace[i]; tmpstr[1] = 0; backref = atoi(tmpstr); /* print out the text held in regmatch[backref] */ if (backref <= matches && regmatch[backref].rm_so != -1) for (j = regmatch[backref].rm_so; j < regmatch[backref].rm_eo; j++) pipeputc(line[j]); } /* if we find a backslash escaped character, print the character */ else if (replace[i] == '\\') { ++i; pipeputc(replace[i]); } /* if we find an unescaped '&' print out the whole matched text. * fortunately, regmatch[0] contains the indicies to the whole matched * expression (kinda seems like it was designed for just such a * purpose...) */ else if (replace[i] == '&' && replace[i-1] != '\\') { int j; for (j = regmatch[0].rm_so; j < regmatch[0].rm_eo; j++) pipeputc(line[j]); } /* nothing special, just print this char of the replacement string to stdout */ else pipeputc(replace[i]); }}static int do_subst_command(const struct sed_cmd *sed_cmd, char **line){ char *hackline = *line; struct pipeline thepipe = { NULL, 0 , 0}; struct pipeline *const pipeline = &thepipe; int altered = 0; regmatch_t *regmatch = NULL; /* we only proceed if the substitution 'search' expression matches */ if (regexec(sed_cmd->sub_match, hackline, 0, NULL, 0) == REG_NOMATCH) return 0; /* whaddaya know, it matched. get the number of back references */ regmatch = xmalloc(sizeof(regmatch_t) * (sed_cmd->num_backrefs+1)); /* allocate more PIPE_GROW bytes if replaced string is larger than original */ thepipe.len = strlen(hackline)+PIPE_GROW; thepipe.buf = xcalloc(1, thepipe.len); /* buffer magic */ thepipe.buf[thepipe.len-1] = PIPE_MAGIC; /* and now, as long as we've got a line to try matching and if we can match * the search string, we make substitutions */ while ((*hackline || !altered) && (regexec(sed_cmd->sub_match, hackline, sed_cmd->num_backrefs+1, regmatch, 0) != REG_NOMATCH) ) { int i; /* print everything before the match */ for (i = 0; i < regmatch[0].rm_so; i++) pipeputc(hackline[i]); /* then print the substitution string */ print_subst_w_backrefs(hackline, sed_cmd->replace, regmatch, pipeline, sed_cmd->num_backrefs); /* advance past the match */ hackline += regmatch[0].rm_eo; /* flag that something has changed */ altered++; /* if we're not doing this globally, get out now */ if (!sed_cmd->sub_g) break; } for (; *hackline; hackline++) pipeputc(*hackline); if (thepipe.buf[thepipe.idx] == PIPE_MAGIC) thepipe.buf[thepipe.idx] = 0; /* cleanup */ free(regmatch); free(*line); *line = thepipe.buf; return altered;}static void process_file(FILE *file){ char *line = NULL; static int linenum = 0; /* GNU sed does not restart counting lines at EOF */ unsigned int still_in_range = 0; int altered; int i; /* go through every line in the file */ while ((line = get_line_from_file(file)) != NULL) { chomp(line); linenum++; altered = 0; /* for every line, go through all the commands */ for (i = 0; i < ncmds; i++) { struct sed_cmd *sed_cmd = &sed_cmds[i]; int deleted = 0; /* * entry point into sedding... */ int matched = ( /* no range necessary */ (sed_cmd->beg_line == 0 && sed_cmd->end_line == 0 && sed_cmd->beg_match == NULL && sed_cmd->end_match == NULL) || /* this line number is the first address we're looking for */ (sed_cmd->beg_line && (sed_cmd->beg_line == linenum)) || /* this line matches our first address regex */ (sed_cmd->beg_match && (regexec(sed_cmd->beg_match, line, 0, NULL, 0) == 0)) || /* we are currently within the beginning & ending address range */ still_in_range ); if (sed_cmd->invert ^ matched) { /* * actual sedding */ switch (sed_cmd->cmd) { case 'p': puts(line); break; case 'd': altered++; deleted = 1; break; case 's': /* * Some special cases for 's' printing to make it compliant with * GNU sed printing behavior (aka "The -n | s///p Matrix"): * * -n ONLY = never print anything regardless of any successful * substitution * * s///p ONLY = always print successful substitutions, even if * the line is going to be printed anyway (line will be printed * twice). * * -n AND s///p = print ONLY a successful substitution ONE TIME; * no other lines are printed - this is the reason why the 'p' * flag exists in the first place. */ /* if the user specified that they didn't want anything printed (i.e., a -n * flag and no 'p' flag after the s///), then there's really no point doing * anything here. */ if (be_quiet && !sed_cmd->sub_p) break; /* we print the line once, unless we were told to be quiet */ if (!be_quiet) altered |= do_subst_command(sed_cmd, &line); /* we also print the line if we were given the 'p' flag * (this is quite possibly the second printing) */ if (sed_cmd->sub_p) altered |= do_subst_command(sed_cmd, &line); if (altered && (i+1 >= ncmds || sed_cmds[i+1].cmd != 's')) puts(line); break; case 'a': puts(line); fputs(sed_cmd->editline, stdout); altered++; break; case 'i': fputs(sed_cmd->editline, stdout); break; case 'c': /* single-address case */ if ((sed_cmd->end_match == NULL && sed_cmd->end_line == 0) /* multi-address case */ /* - matching text */ || (sed_cmd->end_match && (regexec(sed_cmd->end_match, line, 0, NULL, 0) == 0)) /* - matching line numbers */ || (sed_cmd->end_line > 0 && sed_cmd->end_line == linenum)) { fputs(sed_cmd->editline, stdout); } altered++; break; case 'r': { FILE *outfile; puts(line); outfile = fopen(sed_cmd->filename, "r"); if (outfile) print_file(outfile); /* else if we couldn't open the output file, * no biggie, just don't print anything */ altered++; } break; } } /* * exit point from sedding... */ if (matched) { if ( /* this is a single-address command or... */ (sed_cmd->end_line == 0 && sed_cmd->end_match == NULL) || ( /* we were in the middle of our address range (this * isn't the first time through) and.. */ (still_in_range == 1) && ( /* this line number is the last address we're looking for or... */ (sed_cmd->end_line && (sed_cmd->end_line == linenum)) || /* this line matches our last address regex */ (sed_cmd->end_match && (regexec(sed_cmd->end_match, line, 0, NULL, 0) == 0)) ) ) ) { /* we're out of our address range */ still_in_range = 0; } /* didn't hit the exit? then we're still in the middle of an address range */ else { still_in_range = 1; } } if (deleted) break; } /* we will print the line unless we were told to be quiet or if the * line was altered (via a 'd'elete or 's'ubstitution), in which case * the altered line was already printed */ if (!be_quiet && !altered) puts(line); free(line); }}extern int sed_main(int argc, char **argv){ int opt, status = EXIT_SUCCESS;#ifdef CONFIG_FEATURE_CLEAN_UP /* destroy command strings on exit */ if (atexit(destroy_cmd_strs) == -1) perror_msg_and_die("atexit");#endif /* do normal option parsing */ while ((opt = getopt(argc, argv, "ne:f:")) > 0) { switch (opt) { case 'n': be_quiet++; break; case 'e': add_cmd_str(optarg); break; case 'f': load_cmd_file(optarg); break; default: show_usage(); } } /* if we didn't get a pattern from a -e and no command file was specified, * argv[optind] should be the pattern. no pattern, no worky */ if (ncmds == 0) { if (argv[optind] == NULL) show_usage(); else { add_cmd_str(argv[optind]); optind++; } } /* argv[(optind)..(argc-1)] should be names of file to process. If no * files were specified or '-' was specified, take input from stdin. * Otherwise, we process all the files specified. */ if (argv[optind] == NULL || (strcmp(argv[optind], "-") == 0)) { process_file(stdin); } else { int i; FILE *file; for (i = optind; i < argc; i++) { file = wfopen(argv[i], "r"); if (file) { process_file(file); fclose(file); } else status = EXIT_FAILURE; } } return status;}
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -