- typo in documentation
[oweals/busybox.git] / editors / sed.c
index ae8d9832e0b3aaf0651cb72a0502a2138734fe2e..3b0cebda6b3609c5aa36143efe2247005e50cc98 100644 (file)
   resulting sed_cmd_t structures are appended to a linked list
   (sed_cmd_head/sed_cmd_tail).
 
-  process_file() does actual sedding, reading data lines from an input FILE *
+  add_input_file() adds a FILE * to the list of input files.  We need to
+  know them all ahead of time to find the last line for the $ match.
+
+  process_files() does actual sedding, reading data lines from each input FILE *
   (which could be stdin) and applying the sed command list (sed_cmd_head) to
   each of the resulting lines.
 
@@ -54,7 +57,7 @@
         - grouped commands: {cmd1;cmd2}
         - transliteration (y/source-chars/dest-chars/)
         - pattern space hold space storing / swapping (g, h, x)
-        - labels / branching (: label, b, t)
+        - labels / branching (: label, b, t, T)
 
         (Note: Specifying an address (range) to match is *optional*; commands
         default to the whole pattern space if no specific address match was
@@ -62,7 +65,7 @@
 
        Unsupported features:
 
-        - GNU extensions
+        - most GNU extensions
         - and more.
 
        Todo:
 
 #include <stdio.h>
 #include <unistd.h>            /* for getopt() */
-#include <regex.h>
 #include <string.h>            /* for strdup() */
 #include <errno.h>
 #include <ctype.h>             /* for isspace() */
 #include <stdlib.h>
 #include "busybox.h"
+#include "xregex.h"
 
 typedef struct sed_cmd_s {
     /* Ordered by alignment requirements: currently 36 bytes on x86 */
@@ -113,15 +116,18 @@ typedef struct sed_cmd_s {
 /* globals */
 /* options */
 static int be_quiet, in_place, regex_type;
-FILE *nonstdout;
-char *outname,*hold_space;
+static FILE *nonstdout;
+static char *outname,*hold_space;
 
+/* List of input files */
+static int input_file_count,current_input_file;
+static FILE **input_file_list;
 
 static const char bad_format_in_subst[] =
        "bad format in substitution expression";
-const char *const semicolon_whitespace = "; \n\r\t\v";
+static const char *const semicolon_whitespace = "; \n\r\t\v";
 
-regmatch_t regmatch[10];
+static regmatch_t regmatch[10];
 static regex_t *previous_regex_ptr;
 
 /* linked list of sed commands */
@@ -133,7 +139,7 @@ struct append_list {
        char *string;
        struct append_list *next;
 };
-struct append_list *append_head=NULL, *append_tail=NULL;
+static struct append_list *append_head=NULL, *append_tail=NULL;
 
 #ifdef CONFIG_FEATURE_CLEAN_UP
 static void free_and_close_stuff(void)
@@ -171,6 +177,9 @@ static void free_and_close_stuff(void)
        }
 
        if(hold_space) free(hold_space);
+
+    while(current_input_file<input_file_count)
+               fclose(input_file_list[current_input_file++]);
 }
 #endif
 
@@ -431,7 +440,7 @@ static char *parse_cmd_args(sed_cmd_t *sed_cmd, char *cmdstr)
                if(sed_cmd->cmd=='w')
                        sed_cmd->file=bb_xfopen(sed_cmd->string,"w");
        /* handle branch commands */
-       } else if (strchr(":bt", sed_cmd->cmd)) {
+       } else if (strchr(":btT", sed_cmd->cmd)) {
                int length;
 
                while(isspace(*cmdstr)) cmdstr++;
@@ -473,7 +482,7 @@ static char *parse_cmd_args(sed_cmd_t *sed_cmd, char *cmdstr)
 
 /* Parse address+command sets, skipping comment lines. */
 
-void add_cmd(char *cmdstr)
+static void add_cmd(char *cmdstr)
 {
        static char *add_cmd_line=NULL;
        sed_cmd_t *sed_cmd;
@@ -565,7 +574,9 @@ void add_cmd(char *cmdstr)
        }
 }
 
-struct pipeline {
+/* Append to a string, reallocating memory as necessary. */
+
+static struct pipeline {
        char *buf;      /* Space to hold string */
        int idx;        /* Space used */
        int len;        /* Space allocated */
@@ -573,7 +584,7 @@ struct pipeline {
 
 #define PIPE_GROW 64
 
-void pipe_putc(char c)
+static void pipe_putc(char c)
 {
        if(pipeline.idx==pipeline.len) {
                pipeline.buf = xrealloc(pipeline.buf, pipeline.len + PIPE_GROW);
@@ -718,20 +729,29 @@ static void flush_append(void)
        append_head=append_tail=NULL;
 }
 
-/* Get next line of input, flushing append buffer and noting if we hit EOF
- * without a newline on the last line.
+static void add_input_file(FILE *file)
+{
+       input_file_list=xrealloc(input_file_list,(input_file_count+1)*sizeof(FILE *));
+       input_file_list[input_file_count++]=file;
+}
+
+/* Get next line of input from 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(FILE * file, int *no_newline)
+static char *get_next_line(int *no_newline)
 {
-       char *temp;
+       char *temp=NULL;
        int len;
 
        flush_append();
-       temp=bb_get_line_from_file(file);
-       if(temp) {
-               len=strlen(temp);
-               if(len && temp[len-1]=='\n') temp[len-1]=0;
-               else *no_newline=1;
+       while(current_input_file<input_file_count) {
+               temp=bb_get_line_from_file(input_file_list[current_input_file]);
+               if(temp) {
+                       len=strlen(temp);
+                       *no_newline=!(len && temp[len-1]=='\n');
+                       if(!*no_newline) temp[len-1]=0;
+                       break;
+               } else fclose(input_file_list[current_input_file++]);
        }
 
        return temp;
@@ -757,15 +777,15 @@ static int puts_maybe_newline(char *s, FILE *file, int missing_newline, int no_n
 
 #define sed_puts(s,n) missing_newline=puts_maybe_newline(s,nonstdout,missing_newline,n)
 
-static void process_file(FILE *file)
+static void process_files(void)
 {
        char *pattern_space, *next_line;
-       static int linenum = 0, missing_newline=0;
+       int linenum = 0, missing_newline=0;
        int no_newline,next_no_newline=0;
 
-       next_line = get_next_line(file,&next_no_newline);
+       next_line = get_next_line(&next_no_newline);
 
-       /* go through every line in the file */
+       /* go through every line in each file */
        for(;;) {
                sed_cmd_t *sed_cmd;
                int substituted=0;
@@ -775,7 +795,7 @@ static void process_file(FILE *file)
                no_newline=next_no_newline;
 
                /* Read one line in advance so we can act on the last line, the '$' address */
-               next_line = get_next_line(file,&next_no_newline);
+               next_line = get_next_line(&next_no_newline);
                linenum++;
 restart:
                /* for every line, go through all the commands */
@@ -944,7 +964,7 @@ restart:
                                                        free(pattern_space);
                                                        pattern_space = next_line;
                                                        no_newline=next_no_newline;
-                                                       next_line = get_next_line(file,&next_no_newline);
+                                                       next_line = get_next_line(&next_no_newline);
                                                        linenum++;
                                                        break;
                                                }
@@ -974,17 +994,21 @@ restart:
                                                        pattern_space[len]='\n';
                                                        strcpy(pattern_space+len+1, next_line);
                                                        no_newline=next_no_newline;
-                                                       next_line = get_next_line(file,&next_no_newline);
+                                                       next_line = get_next_line(&next_no_newline);
                                                        linenum++;
                                                }
                                                break;
                                        }
 
-                                       /* Test if substition worked, branch if so. */
+                                       /* Test/branch if substitution occurred */
                                        case 't':
-                                               if (!substituted) break;
+                                               if(!substituted) break;
                                                substituted=0;
-                                                       /* Fall through */
+                                               /* Fall through */
+                                       /* Test/branch if substitution didn't occur */
+                                       case 'T':
+                                               if (substituted) break;
+                                               /* Fall through */
                                        /* Branch to label */
                                        case 'b':
                                                if (!sed_cmd->string) goto discard_commands;
@@ -1001,6 +1025,7 @@ restart:
                                                        for (j = 0; sed_cmd->string[j]; j += 2) {
                                                                if (pattern_space[i] == sed_cmd->string[j]) {
                                                                        pattern_space[i] = sed_cmd->string[j + 1];
+                                                                       break;
                                                                }
                                                        }
                                                }
@@ -1150,8 +1175,7 @@ extern int sed_main(int argc, char **argv)
                }
        }
 
-       /* 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 we didn't get a pattern from -e or -f, use argv[optind] */
        if(getpat) {
                if (argv[optind] == NULL)
                        bb_show_usage();
@@ -1168,49 +1192,47 @@ extern int sed_main(int argc, char **argv)
         * files were specified or '-' was specified, take input from stdin.
         * Otherwise, we process all the files specified. */
        if (argv[optind] == NULL) {
-               if(in_place) {
-                       fprintf(stderr,"sed: Filename required for -i\n");
-                       exit(1);
-               }
-               process_file(stdin);
+               if(in_place) bb_error_msg_and_die("Filename required for -i");
+               add_input_file(stdin);
+               process_files();
        } else {
                int i;
                FILE *file;
 
                for (i = optind; i < argc; i++) {
                        if(!strcmp(argv[i], "-") && !in_place) {
-                               process_file(stdin);
+                               add_input_file(stdin);
+                               process_files();
                        } else {
                                file = bb_wfopen(argv[i], "r");
                                if (file) {
                                        if(in_place) {
                                                struct stat statbuf;
+                                               int nonstdoutfd;
+                                               
                                                outname=bb_xstrndup(argv[i],strlen(argv[i])+6);
                                                strcat(outname,"XXXXXX");
+                                               if(-1==(nonstdoutfd=mkstemp(outname)))
+                                                       bb_error_msg_and_die("no temp file");
+                                               nonstdout=fdopen(nonstdoutfd,"w");
                                                /* Set permissions of output file */
                                                fstat(fileno(file),&statbuf);
-                                               mkstemp(outname);
-                                               nonstdout=bb_wfopen(outname,"w");
-                                               /* Set permissions of output file */
-                                               fstat(fileno(file),&statbuf);
-                                               fchmod(fileno(nonstdout),statbuf.st_mode);
-                                               atexit(cleanup_outname);
-                                       }
-                                       process_file(file);
-                                       fclose(file);
-                                       if(in_place) {
+                                               fchmod(nonstdoutfd,statbuf.st_mode);
+                                               add_input_file(file);
+                                               process_files();
                                                fclose(nonstdout);
                                                nonstdout=stdout;
                                                unlink(argv[i]);
                                                rename(outname,argv[i]);
                                                free(outname);
                                                outname=0;
-                                       }
+                                       } else add_input_file(file);
                                } else {
                                        status = EXIT_FAILURE;
                                }
                        }
                }
+               if(input_file_count>current_input_file) process_files();
        }
 
        return status;