19318a62d7
It contains changes pushed by pstef@ when working on improving the completion for /bin/sh
858 lines
21 KiB
C
858 lines
21 KiB
C
/* $NetBSD: filecomplete.c,v 1.67 2021/03/28 13:39:39 christos Exp $ */
|
|
|
|
/*-
|
|
* Copyright (c) 1997 The NetBSD Foundation, Inc.
|
|
* All rights reserved.
|
|
*
|
|
* This code is derived from software contributed to The NetBSD Foundation
|
|
* by Jaromir Dolecek.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
* documentation and/or other materials provided with the distribution.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
|
|
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
|
|
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
|
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
|
|
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
|
|
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
|
|
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
|
|
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
|
|
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
|
|
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
* POSSIBILITY OF SUCH DAMAGE.
|
|
*/
|
|
|
|
#include "config.h"
|
|
#if !defined(lint) && !defined(SCCSID)
|
|
__RCSID("$NetBSD: filecomplete.c,v 1.67 2021/03/28 13:39:39 christos Exp $");
|
|
#endif /* not lint && not SCCSID */
|
|
|
|
#include <sys/types.h>
|
|
#include <sys/stat.h>
|
|
#include <dirent.h>
|
|
#include <errno.h>
|
|
#include <fcntl.h>
|
|
#include <limits.h>
|
|
#include <pwd.h>
|
|
#include <stdio.h>
|
|
#include <stdlib.h>
|
|
#include <string.h>
|
|
#include <unistd.h>
|
|
|
|
#include "el.h"
|
|
#include "filecomplete.h"
|
|
|
|
static const wchar_t break_chars[] = L" \t\n\"\\'`@$><=;|&{(";
|
|
|
|
/********************************/
|
|
/* completion functions */
|
|
|
|
/*
|
|
* does tilde expansion of strings of type ``~user/foo''
|
|
* if ``user'' isn't valid user name or ``txt'' doesn't start
|
|
* w/ '~', returns pointer to strdup()ed copy of ``txt''
|
|
*
|
|
* it's the caller's responsibility to free() the returned string
|
|
*/
|
|
char *
|
|
fn_tilde_expand(const char *txt)
|
|
{
|
|
#if defined(HAVE_GETPW_R_POSIX) || defined(HAVE_GETPW_R_DRAFT)
|
|
struct passwd pwres;
|
|
char pwbuf[1024];
|
|
#endif
|
|
struct passwd *pass;
|
|
char *temp;
|
|
size_t len = 0;
|
|
|
|
if (txt[0] != '~')
|
|
return strdup(txt);
|
|
|
|
temp = strchr(txt + 1, '/');
|
|
if (temp == NULL) {
|
|
temp = strdup(txt + 1);
|
|
if (temp == NULL)
|
|
return NULL;
|
|
} else {
|
|
/* text until string after slash */
|
|
len = (size_t)(temp - txt + 1);
|
|
temp = el_calloc(len, sizeof(*temp));
|
|
if (temp == NULL)
|
|
return NULL;
|
|
(void)strlcpy(temp, txt + 1, len - 1);
|
|
}
|
|
if (temp[0] == 0) {
|
|
#ifdef HAVE_GETPW_R_POSIX
|
|
if (getpwuid_r(getuid(), &pwres, pwbuf, sizeof(pwbuf),
|
|
&pass) != 0)
|
|
pass = NULL;
|
|
#elif HAVE_GETPW_R_DRAFT
|
|
pass = getpwuid_r(getuid(), &pwres, pwbuf, sizeof(pwbuf));
|
|
#else
|
|
pass = getpwuid(getuid());
|
|
#endif
|
|
} else {
|
|
#ifdef HAVE_GETPW_R_POSIX
|
|
if (getpwnam_r(temp, &pwres, pwbuf, sizeof(pwbuf), &pass) != 0)
|
|
pass = NULL;
|
|
#elif HAVE_GETPW_R_DRAFT
|
|
pass = getpwnam_r(temp, &pwres, pwbuf, sizeof(pwbuf));
|
|
#else
|
|
pass = getpwnam(temp);
|
|
#endif
|
|
}
|
|
el_free(temp); /* value no more needed */
|
|
if (pass == NULL)
|
|
return strdup(txt);
|
|
|
|
/* update pointer txt to point at string immedially following */
|
|
/* first slash */
|
|
txt += len;
|
|
|
|
len = strlen(pass->pw_dir) + 1 + strlen(txt) + 1;
|
|
temp = el_calloc(len, sizeof(*temp));
|
|
if (temp == NULL)
|
|
return NULL;
|
|
(void)snprintf(temp, len, "%s/%s", pass->pw_dir, txt);
|
|
|
|
return temp;
|
|
}
|
|
|
|
static int
|
|
needs_escaping(char c)
|
|
{
|
|
switch (c) {
|
|
case '\'':
|
|
case '"':
|
|
case '(':
|
|
case ')':
|
|
case '\\':
|
|
case '<':
|
|
case '>':
|
|
case '$':
|
|
case '#':
|
|
case ' ':
|
|
case '\n':
|
|
case '\t':
|
|
case '?':
|
|
case ';':
|
|
case '`':
|
|
case '@':
|
|
case '=':
|
|
case '|':
|
|
case '{':
|
|
case '}':
|
|
case '&':
|
|
case '*':
|
|
case '[':
|
|
return 1;
|
|
default:
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
static int
|
|
needs_dquote_escaping(char c)
|
|
{
|
|
switch (c) {
|
|
case '"':
|
|
case '\\':
|
|
case '`':
|
|
case '$':
|
|
return 1;
|
|
default:
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
|
|
static wchar_t *
|
|
unescape_string(const wchar_t *string, size_t length)
|
|
{
|
|
size_t i;
|
|
size_t j = 0;
|
|
wchar_t *unescaped = el_calloc(length + 1, sizeof(*string));
|
|
if (unescaped == NULL)
|
|
return NULL;
|
|
for (i = 0; i < length ; i++) {
|
|
if (string[i] == '\\')
|
|
continue;
|
|
unescaped[j++] = string[i];
|
|
}
|
|
unescaped[j] = 0;
|
|
return unescaped;
|
|
}
|
|
|
|
static char *
|
|
escape_filename(EditLine * el, const char *filename, int single_match,
|
|
const char *(*app_func)(const char *))
|
|
{
|
|
size_t original_len = 0;
|
|
size_t escaped_character_count = 0;
|
|
size_t offset = 0;
|
|
size_t newlen;
|
|
const char *s;
|
|
char c;
|
|
size_t s_quoted = 0; /* does the input contain a single quote */
|
|
size_t d_quoted = 0; /* does the input contain a double quote */
|
|
char *escaped_str;
|
|
wchar_t *temp = el->el_line.buffer;
|
|
const char *append_char = NULL;
|
|
|
|
if (filename == NULL)
|
|
return NULL;
|
|
|
|
while (temp != el->el_line.cursor) {
|
|
/*
|
|
* If we see a single quote but have not seen a double quote
|
|
* so far set/unset s_quote
|
|
*/
|
|
if (temp[0] == '\'' && !d_quoted)
|
|
s_quoted = !s_quoted;
|
|
/*
|
|
* vice versa to the above condition
|
|
*/
|
|
else if (temp[0] == '"' && !s_quoted)
|
|
d_quoted = !d_quoted;
|
|
temp++;
|
|
}
|
|
|
|
/* Count number of special characters so that we can calculate
|
|
* number of extra bytes needed in the new string
|
|
*/
|
|
for (s = filename; *s; s++, original_len++) {
|
|
c = *s;
|
|
/* Inside a single quote only single quotes need escaping */
|
|
if (s_quoted && c == '\'') {
|
|
escaped_character_count += 3;
|
|
continue;
|
|
}
|
|
/* Inside double quotes only ", \, ` and $ need escaping */
|
|
if (d_quoted && needs_dquote_escaping(c)) {
|
|
escaped_character_count++;
|
|
continue;
|
|
}
|
|
if (!s_quoted && !d_quoted && needs_escaping(c))
|
|
escaped_character_count++;
|
|
}
|
|
|
|
newlen = original_len + escaped_character_count + 1;
|
|
if (s_quoted || d_quoted)
|
|
newlen++;
|
|
|
|
if (single_match && app_func)
|
|
newlen++;
|
|
|
|
if ((escaped_str = el_malloc(newlen)) == NULL)
|
|
return NULL;
|
|
|
|
for (s = filename; *s; s++) {
|
|
c = *s;
|
|
if (!needs_escaping(c)) {
|
|
/* no escaping is required continue as usual */
|
|
escaped_str[offset++] = c;
|
|
continue;
|
|
}
|
|
|
|
/* single quotes inside single quotes require special handling */
|
|
if (c == '\'' && s_quoted) {
|
|
escaped_str[offset++] = '\'';
|
|
escaped_str[offset++] = '\\';
|
|
escaped_str[offset++] = '\'';
|
|
escaped_str[offset++] = '\'';
|
|
continue;
|
|
}
|
|
|
|
/* Otherwise no escaping needed inside single quotes */
|
|
if (s_quoted) {
|
|
escaped_str[offset++] = c;
|
|
continue;
|
|
}
|
|
|
|
/* No escaping needed inside a double quoted string either
|
|
* unless we see a '$', '\', '`', or '"' (itself)
|
|
*/
|
|
if (d_quoted && !needs_dquote_escaping(c)) {
|
|
escaped_str[offset++] = c;
|
|
continue;
|
|
}
|
|
|
|
/* If we reach here that means escaping is actually needed */
|
|
escaped_str[offset++] = '\\';
|
|
escaped_str[offset++] = c;
|
|
}
|
|
|
|
if (single_match && app_func) {
|
|
escaped_str[offset] = 0;
|
|
append_char = app_func(filename);
|
|
/* we want to append space only if we are not inside quotes */
|
|
if (append_char[0] == ' ') {
|
|
if (!s_quoted && !d_quoted)
|
|
escaped_str[offset++] = append_char[0];
|
|
} else
|
|
escaped_str[offset++] = append_char[0];
|
|
}
|
|
|
|
/* close the quotes if single match and the match is not a directory */
|
|
if (single_match && (append_char && append_char[0] == ' ')) {
|
|
if (s_quoted)
|
|
escaped_str[offset++] = '\'';
|
|
else if (d_quoted)
|
|
escaped_str[offset++] = '"';
|
|
}
|
|
|
|
escaped_str[offset] = 0;
|
|
return escaped_str;
|
|
}
|
|
|
|
/*
|
|
* return first found file name starting by the ``text'' or NULL if no
|
|
* such file can be found
|
|
* value of ``state'' is ignored
|
|
*
|
|
* it's the caller's responsibility to free the returned string
|
|
*/
|
|
char *
|
|
fn_filename_completion_function(const char *text, int state)
|
|
{
|
|
static DIR *dir = NULL;
|
|
static char *filename = NULL, *dirname = NULL, *dirpath = NULL;
|
|
static size_t filename_len = 0;
|
|
struct dirent *entry;
|
|
char *temp;
|
|
size_t len;
|
|
|
|
if (state == 0 || dir == NULL) {
|
|
temp = strrchr(text, '/');
|
|
if (temp) {
|
|
char *nptr;
|
|
temp++;
|
|
nptr = el_realloc(filename, (strlen(temp) + 1) *
|
|
sizeof(*nptr));
|
|
if (nptr == NULL) {
|
|
el_free(filename);
|
|
filename = NULL;
|
|
return NULL;
|
|
}
|
|
filename = nptr;
|
|
(void)strcpy(filename, temp);
|
|
len = (size_t)(temp - text); /* including last slash */
|
|
|
|
nptr = el_realloc(dirname, (len + 1) *
|
|
sizeof(*nptr));
|
|
if (nptr == NULL) {
|
|
el_free(dirname);
|
|
dirname = NULL;
|
|
return NULL;
|
|
}
|
|
dirname = nptr;
|
|
(void)strlcpy(dirname, text, len + 1);
|
|
} else {
|
|
el_free(filename);
|
|
if (*text == 0)
|
|
filename = NULL;
|
|
else {
|
|
filename = strdup(text);
|
|
if (filename == NULL)
|
|
return NULL;
|
|
}
|
|
el_free(dirname);
|
|
dirname = NULL;
|
|
}
|
|
|
|
if (dir != NULL) {
|
|
(void)closedir(dir);
|
|
dir = NULL;
|
|
}
|
|
|
|
/* support for ``~user'' syntax */
|
|
|
|
el_free(dirpath);
|
|
dirpath = NULL;
|
|
if (dirname == NULL) {
|
|
if ((dirname = strdup("")) == NULL)
|
|
return NULL;
|
|
dirpath = strdup("./");
|
|
} else if (*dirname == '~')
|
|
dirpath = fn_tilde_expand(dirname);
|
|
else
|
|
dirpath = strdup(dirname);
|
|
|
|
if (dirpath == NULL)
|
|
return NULL;
|
|
|
|
dir = opendir(dirpath);
|
|
if (!dir)
|
|
return NULL; /* cannot open the directory */
|
|
|
|
/* will be used in cycle */
|
|
filename_len = filename ? strlen(filename) : 0;
|
|
}
|
|
|
|
/* find the match */
|
|
while ((entry = readdir(dir)) != NULL) {
|
|
/* skip . and .. */
|
|
if (entry->d_name[0] == '.' && (!entry->d_name[1]
|
|
|| (entry->d_name[1] == '.' && !entry->d_name[2])))
|
|
continue;
|
|
if (filename_len == 0)
|
|
break;
|
|
/* otherwise, get first entry where first */
|
|
/* filename_len characters are equal */
|
|
if (entry->d_name[0] == filename[0]
|
|
#if HAVE_STRUCT_DIRENT_D_NAMLEN
|
|
&& entry->d_namlen >= filename_len
|
|
#else
|
|
&& strlen(entry->d_name) >= filename_len
|
|
#endif
|
|
&& strncmp(entry->d_name, filename,
|
|
filename_len) == 0)
|
|
break;
|
|
}
|
|
|
|
if (entry) { /* match found */
|
|
|
|
#if HAVE_STRUCT_DIRENT_D_NAMLEN
|
|
len = entry->d_namlen;
|
|
#else
|
|
len = strlen(entry->d_name);
|
|
#endif
|
|
|
|
len = strlen(dirname) + len + 1;
|
|
temp = el_calloc(len, sizeof(*temp));
|
|
if (temp == NULL)
|
|
return NULL;
|
|
(void)snprintf(temp, len, "%s%s", dirname, entry->d_name);
|
|
} else {
|
|
(void)closedir(dir);
|
|
dir = NULL;
|
|
temp = NULL;
|
|
}
|
|
|
|
return temp;
|
|
}
|
|
|
|
|
|
static const char *
|
|
append_char_function(const char *name)
|
|
{
|
|
struct stat stbuf;
|
|
char *expname = *name == '~' ? fn_tilde_expand(name) : NULL;
|
|
const char *rs = " ";
|
|
|
|
if (stat(expname ? expname : name, &stbuf) == -1)
|
|
goto out;
|
|
if (S_ISDIR(stbuf.st_mode))
|
|
rs = "/";
|
|
out:
|
|
if (expname)
|
|
el_free(expname);
|
|
return rs;
|
|
}
|
|
/*
|
|
* returns list of completions for text given
|
|
* non-static for readline.
|
|
*/
|
|
char ** completion_matches(const char *, char *(*)(const char *, int));
|
|
char **
|
|
completion_matches(const char *text, char *(*genfunc)(const char *, int))
|
|
{
|
|
char **match_list = NULL, *retstr, *prevstr;
|
|
size_t match_list_len, max_equal, which, i;
|
|
size_t matches;
|
|
|
|
matches = 0;
|
|
match_list_len = 1;
|
|
while ((retstr = (*genfunc) (text, (int)matches)) != NULL) {
|
|
/* allow for list terminator here */
|
|
if (matches + 3 >= match_list_len) {
|
|
char **nmatch_list;
|
|
while (matches + 3 >= match_list_len)
|
|
match_list_len <<= 1;
|
|
nmatch_list = el_realloc(match_list,
|
|
match_list_len * sizeof(*nmatch_list));
|
|
if (nmatch_list == NULL) {
|
|
el_free(match_list);
|
|
return NULL;
|
|
}
|
|
match_list = nmatch_list;
|
|
|
|
}
|
|
match_list[++matches] = retstr;
|
|
}
|
|
|
|
if (!match_list)
|
|
return NULL; /* nothing found */
|
|
|
|
/* find least denominator and insert it to match_list[0] */
|
|
which = 2;
|
|
prevstr = match_list[1];
|
|
max_equal = strlen(prevstr);
|
|
for (; which <= matches; which++) {
|
|
for (i = 0; i < max_equal &&
|
|
prevstr[i] == match_list[which][i]; i++)
|
|
continue;
|
|
max_equal = i;
|
|
}
|
|
|
|
retstr = el_calloc(max_equal + 1, sizeof(*retstr));
|
|
if (retstr == NULL) {
|
|
el_free(match_list);
|
|
return NULL;
|
|
}
|
|
(void)strlcpy(retstr, match_list[1], max_equal + 1);
|
|
match_list[0] = retstr;
|
|
|
|
/* add NULL as last pointer to the array */
|
|
match_list[matches + 1] = NULL;
|
|
|
|
return match_list;
|
|
}
|
|
|
|
/*
|
|
* Sort function for qsort(). Just wrapper around strcasecmp().
|
|
*/
|
|
static int
|
|
_fn_qsort_string_compare(const void *i1, const void *i2)
|
|
{
|
|
const char *s1 = ((const char * const *)i1)[0];
|
|
const char *s2 = ((const char * const *)i2)[0];
|
|
|
|
return strcasecmp(s1, s2);
|
|
}
|
|
|
|
/*
|
|
* Display list of strings in columnar format on readline's output stream.
|
|
* 'matches' is list of strings, 'num' is number of strings in 'matches',
|
|
* 'width' is maximum length of string in 'matches'.
|
|
*
|
|
* matches[0] is not one of the match strings, but it is counted in
|
|
* num, so the strings are matches[1] *through* matches[num-1].
|
|
*/
|
|
void
|
|
fn_display_match_list(EditLine * el, char **matches, size_t num, size_t width,
|
|
const char *(*app_func) (const char *))
|
|
{
|
|
size_t line, lines, col, cols, thisguy;
|
|
int screenwidth = el->el_terminal.t_size.h;
|
|
if (app_func == NULL)
|
|
app_func = append_char_function;
|
|
|
|
/* Ignore matches[0]. Avoid 1-based array logic below. */
|
|
matches++;
|
|
num--;
|
|
|
|
/*
|
|
* Find out how many entries can be put on one line; count
|
|
* with one space between strings the same way it's printed.
|
|
*/
|
|
cols = (size_t)screenwidth / (width + 2);
|
|
if (cols == 0)
|
|
cols = 1;
|
|
|
|
/* how many lines of output, rounded up */
|
|
lines = (num + cols - 1) / cols;
|
|
|
|
/* Sort the items. */
|
|
qsort(matches, num, sizeof(char *), _fn_qsort_string_compare);
|
|
|
|
/*
|
|
* On the ith line print elements i, i+lines, i+lines*2, etc.
|
|
*/
|
|
for (line = 0; line < lines; line++) {
|
|
for (col = 0; col < cols; col++) {
|
|
thisguy = line + col * lines;
|
|
if (thisguy >= num)
|
|
break;
|
|
(void)fprintf(el->el_outfile, "%s%s%s",
|
|
col == 0 ? "" : " ", matches[thisguy],
|
|
(*app_func)(matches[thisguy]));
|
|
(void)fprintf(el->el_outfile, "%-*s",
|
|
(int) (width - strlen(matches[thisguy])), "");
|
|
}
|
|
(void)fprintf(el->el_outfile, "\n");
|
|
}
|
|
}
|
|
|
|
static wchar_t *
|
|
find_word_to_complete(const wchar_t * cursor, const wchar_t * buffer,
|
|
const wchar_t * word_break, const wchar_t * special_prefixes, size_t * length,
|
|
int do_unescape)
|
|
{
|
|
/* We now look backwards for the start of a filename/variable word */
|
|
const wchar_t *ctemp = cursor;
|
|
wchar_t *temp;
|
|
size_t len;
|
|
|
|
/* if the cursor is placed at a slash or a quote, we need to find the
|
|
* word before it
|
|
*/
|
|
if (ctemp > buffer) {
|
|
switch (ctemp[-1]) {
|
|
case '\\':
|
|
case '\'':
|
|
case '"':
|
|
ctemp--;
|
|
break;
|
|
default:
|
|
break;
|
|
}
|
|
}
|
|
|
|
for (;;) {
|
|
if (ctemp <= buffer)
|
|
break;
|
|
if (wcschr(word_break, ctemp[-1])) {
|
|
if (ctemp - buffer >= 2 && ctemp[-2] == '\\') {
|
|
ctemp -= 2;
|
|
continue;
|
|
}
|
|
break;
|
|
}
|
|
if (special_prefixes && wcschr(special_prefixes, ctemp[-1]))
|
|
break;
|
|
ctemp--;
|
|
}
|
|
|
|
len = (size_t) (cursor - ctemp);
|
|
if (len == 1 && (ctemp[0] == '\'' || ctemp[0] == '"')) {
|
|
len = 0;
|
|
ctemp++;
|
|
}
|
|
*length = len;
|
|
if (do_unescape) {
|
|
wchar_t *unescaped_word = unescape_string(ctemp, len);
|
|
if (unescaped_word == NULL)
|
|
return NULL;
|
|
return unescaped_word;
|
|
}
|
|
temp = el_malloc((len + 1) * sizeof(*temp));
|
|
(void) wcsncpy(temp, ctemp, len);
|
|
temp[len] = '\0';
|
|
return temp;
|
|
}
|
|
|
|
/*
|
|
* Complete the word at or before point,
|
|
* 'what_to_do' says what to do with the completion.
|
|
* \t means do standard completion.
|
|
* `?' means list the possible completions.
|
|
* `*' means insert all of the possible completions.
|
|
* `!' means to do standard completion, and list all possible completions if
|
|
* there is more than one.
|
|
*
|
|
* Note: '*' support is not implemented
|
|
* '!' could never be invoked
|
|
*/
|
|
int
|
|
fn_complete2(EditLine *el,
|
|
char *(*complete_func)(const char *, int),
|
|
char **(*attempted_completion_function)(const char *, int, int),
|
|
const wchar_t *word_break, const wchar_t *special_prefixes,
|
|
const char *(*app_func)(const char *), size_t query_items,
|
|
int *completion_type, int *over, int *point, int *end,
|
|
unsigned int flags)
|
|
{
|
|
const LineInfoW *li;
|
|
wchar_t *temp;
|
|
char **matches;
|
|
char *completion;
|
|
size_t len;
|
|
int what_to_do = '\t';
|
|
int retval = CC_NORM;
|
|
int do_unescape = flags & FN_QUOTE_MATCH;
|
|
|
|
if (el->el_state.lastcmd == el->el_state.thiscmd)
|
|
what_to_do = '?';
|
|
|
|
/* readline's rl_complete() has to be told what we did... */
|
|
if (completion_type != NULL)
|
|
*completion_type = what_to_do;
|
|
|
|
if (!complete_func)
|
|
complete_func = fn_filename_completion_function;
|
|
if (!app_func)
|
|
app_func = append_char_function;
|
|
|
|
li = el_wline(el);
|
|
temp = find_word_to_complete(li->cursor,
|
|
li->buffer, word_break, special_prefixes, &len, do_unescape);
|
|
if (temp == NULL)
|
|
goto out;
|
|
|
|
/* these can be used by function called in completion_matches() */
|
|
/* or (*attempted_completion_function)() */
|
|
if (point != NULL)
|
|
*point = (int)(li->cursor - li->buffer);
|
|
if (end != NULL)
|
|
*end = (int)(li->lastchar - li->buffer);
|
|
|
|
if (attempted_completion_function) {
|
|
int cur_off = (int)(li->cursor - li->buffer);
|
|
matches = (*attempted_completion_function)(
|
|
ct_encode_string(temp, &el->el_scratch),
|
|
cur_off - (int)len, cur_off);
|
|
} else
|
|
matches = NULL;
|
|
if (!attempted_completion_function ||
|
|
(over != NULL && !*over && !matches))
|
|
matches = completion_matches(
|
|
ct_encode_string(temp, &el->el_scratch), complete_func);
|
|
|
|
if (over != NULL)
|
|
*over = 0;
|
|
|
|
if (matches == NULL) {
|
|
goto out;
|
|
}
|
|
int i;
|
|
size_t matches_num, maxlen, match_len, match_display=1;
|
|
int single_match = matches[2] == NULL &&
|
|
(matches[1] == NULL || strcmp(matches[0], matches[1]) == 0);
|
|
|
|
retval = CC_REFRESH;
|
|
|
|
if (matches[0][0] != '\0') {
|
|
el_deletestr(el, (int)len);
|
|
if (flags & FN_QUOTE_MATCH)
|
|
completion = escape_filename(el, matches[0],
|
|
single_match, app_func);
|
|
else
|
|
completion = strdup(matches[0]);
|
|
if (completion == NULL)
|
|
goto out;
|
|
|
|
/*
|
|
* Replace the completed string with the common part of
|
|
* all possible matches if there is a possible completion.
|
|
*/
|
|
el_winsertstr(el,
|
|
ct_decode_string(completion, &el->el_scratch));
|
|
|
|
if (single_match && attempted_completion_function &&
|
|
!(flags & FN_QUOTE_MATCH))
|
|
{
|
|
/*
|
|
* We found an exact match. Add a space after
|
|
* it, unless we do filename completion and the
|
|
* object is a directory. Also do necessary
|
|
* escape quoting
|
|
*/
|
|
el_winsertstr(el, ct_decode_string(
|
|
(*app_func)(completion), &el->el_scratch));
|
|
}
|
|
free(completion);
|
|
}
|
|
|
|
|
|
if (!single_match && (what_to_do == '!' || what_to_do == '?')) {
|
|
/*
|
|
* More than one match and requested to list possible
|
|
* matches.
|
|
*/
|
|
|
|
for(i = 1, maxlen = 0; matches[i]; i++) {
|
|
match_len = strlen(matches[i]);
|
|
if (match_len > maxlen)
|
|
maxlen = match_len;
|
|
}
|
|
/* matches[1] through matches[i-1] are available */
|
|
matches_num = (size_t)(i - 1);
|
|
|
|
/* newline to get on next line from command line */
|
|
(void)fprintf(el->el_outfile, "\n");
|
|
|
|
/*
|
|
* If there are too many items, ask user for display
|
|
* confirmation.
|
|
*/
|
|
if (matches_num > query_items) {
|
|
(void)fprintf(el->el_outfile,
|
|
"Display all %zu possibilities? (y or n) ",
|
|
matches_num);
|
|
(void)fflush(el->el_outfile);
|
|
if (getc(stdin) != 'y')
|
|
match_display = 0;
|
|
(void)fprintf(el->el_outfile, "\n");
|
|
}
|
|
|
|
if (match_display) {
|
|
/*
|
|
* Interface of this function requires the
|
|
* strings be matches[1..num-1] for compat.
|
|
* We have matches_num strings not counting
|
|
* the prefix in matches[0], so we need to
|
|
* add 1 to matches_num for the call.
|
|
*/
|
|
fn_display_match_list(el, matches,
|
|
matches_num+1, maxlen, app_func);
|
|
}
|
|
retval = CC_REDISPLAY;
|
|
} else if (matches[0][0]) {
|
|
/*
|
|
* There was some common match, but the name was
|
|
* not complete enough. Next tab will print possible
|
|
* completions.
|
|
*/
|
|
el_beep(el);
|
|
} else {
|
|
/* lcd is not a valid object - further specification */
|
|
/* is needed */
|
|
el_beep(el);
|
|
retval = CC_NORM;
|
|
}
|
|
|
|
/* free elements of array and the array itself */
|
|
for (i = 0; matches[i]; i++)
|
|
el_free(matches[i]);
|
|
el_free(matches);
|
|
matches = NULL;
|
|
|
|
out:
|
|
el_free(temp);
|
|
return retval;
|
|
}
|
|
|
|
int
|
|
fn_complete(EditLine *el,
|
|
char *(*complete_func)(const char *, int),
|
|
char **(*attempted_completion_function)(const char *, int, int),
|
|
const wchar_t *word_break, const wchar_t *special_prefixes,
|
|
const char *(*app_func)(const char *), size_t query_items,
|
|
int *completion_type, int *over, int *point, int *end)
|
|
{
|
|
return fn_complete2(el, complete_func, attempted_completion_function,
|
|
word_break, special_prefixes, app_func, query_items,
|
|
completion_type, over, point, end,
|
|
attempted_completion_function ? 0 : FN_QUOTE_MATCH);
|
|
}
|
|
|
|
/*
|
|
* el-compatible wrapper around rl_complete; needed for key binding
|
|
*/
|
|
/* ARGSUSED */
|
|
unsigned char
|
|
_el_fn_complete(EditLine *el, int ch __attribute__((__unused__)))
|
|
{
|
|
return (unsigned char)fn_complete(el, NULL, NULL,
|
|
break_chars, NULL, NULL, (size_t)100,
|
|
NULL, NULL, NULL, NULL);
|
|
}
|
|
|
|
/*
|
|
* el-compatible wrapper around rl_complete; needed for key binding
|
|
*/
|
|
/* ARGSUSED */
|
|
unsigned char
|
|
_el_fn_sh_complete(EditLine *el, int ch)
|
|
{
|
|
return _el_fn_complete(el, ch);
|
|
}
|