This is the mail archive of the
gdb-patches@sources.redhat.com
mailing list for the GDB project.
Re: Patch: complete -vs- duplicates, take 2
- From: Tom Tromey <tromey at redhat dot com>
- To: Eli Zaretskii <eliz at is dot elta dot co dot il>
- Cc: fnasser at redhat dot com, gdb-patches at sources dot redhat dot com
- Date: 14 Feb 2002 10:51:14 -0700
- Subject: Re: Patch: complete -vs- duplicates, take 2
- References: <87666g1nws.fsf@creche.redhat.com> <3C517C33.13B8A9D4@redhat.com> <2950-Fri25Jan2002204302+0200-eliz@is.elta.co.il>
- Reply-to: tromey at redhat dot com
>>>>> "Eli" == Eli Zaretskii <eliz@is.elta.co.il> writes:
Eli> Tom, could you please play with different values of context line
Eli> numbers (the NUM parameter in "diff -U NUM"), to produce diffs
Eli> for completer.c that would show the real changes, and post just
Eli> that part of the patch?
I finally found time to look into this. I think the reformatting will
never be pretty due to reindentation. I tried several (more than 10)
combinations of -U and --horizon-lines and never got anything really
readable.
So instead I've simply posted the entire functions in question.
Basically complete_line() was made by cut-and-paste from
line_completion_function(). No semantics were changed.
Tom
/* Generate completions all at once. Returns a NULL-terminated array
of strings. Both the array and each element are allocated with
xmalloc. It can also return NULL if there are no completions.
TEXT is the caller's idea of the "word" we are looking at.
LINE_BUFFER is available to be looked at; it contains the entire text
of the line. POINT is the offset in that line of the cursor. You
should pretend that the line ends at POINT. */
char **
complete_line (char *text, char *line_buffer, int point)
{
char **list = NULL;
char *tmp_command, *p;
/* Pointer within tmp_command which corresponds to text. */
char *word;
struct cmd_list_element *c, *result_list;
/* Choose the default set of word break characters to break completions.
If we later find out that we are doing completions on command strings
(as opposed to strings supplied by the individual command completer
functions, which can be any string) then we will switch to the
special word break set for command strings, which leaves out the
'-' character used in some commands. */
rl_completer_word_break_characters =
gdb_completer_word_break_characters;
/* Decide whether to complete on a list of gdb commands or on symbols. */
tmp_command = (char *) alloca (point + 1);
p = tmp_command;
strncpy (tmp_command, line_buffer, point);
tmp_command[point] = '\0';
/* Since text always contains some number of characters leading up
to point, we can find the equivalent position in tmp_command
by subtracting that many characters from the end of tmp_command. */
word = tmp_command + point - strlen (text);
if (point == 0)
{
/* An empty line we want to consider ambiguous; that is, it
could be any command. */
c = (struct cmd_list_element *) -1;
result_list = 0;
}
else
{
c = lookup_cmd_1 (&p, cmdlist, &result_list, 1);
}
/* Move p up to the next interesting thing. */
while (*p == ' ' || *p == '\t')
{
p++;
}
if (!c)
{
/* It is an unrecognized command. So there are no
possible completions. */
list = NULL;
}
else if (c == (struct cmd_list_element *) -1)
{
char *q;
/* lookup_cmd_1 advances p up to the first ambiguous thing, but
doesn't advance over that thing itself. Do so now. */
q = p;
while (*q && (isalnum (*q) || *q == '-' || *q == '_'))
++q;
if (q != tmp_command + point)
{
/* There is something beyond the ambiguous
command, so there are no possible completions. For
example, "info t " or "info t foo" does not complete
to anything, because "info t" can be "info target" or
"info terminal". */
list = NULL;
}
else
{
/* We're trying to complete on the command which was ambiguous.
This we can deal with. */
if (result_list)
{
list = complete_on_cmdlist (*result_list->prefixlist, p,
word);
}
else
{
list = complete_on_cmdlist (cmdlist, p, word);
}
/* Insure that readline does the right thing with respect to
inserting quotes. */
rl_completer_word_break_characters =
gdb_completer_command_word_break_characters;
}
}
else
{
/* We've recognized a full command. */
if (p == tmp_command + point)
{
/* There is no non-whitespace in the line beyond the command. */
if (p[-1] == ' ' || p[-1] == '\t')
{
/* The command is followed by whitespace; we need to complete
on whatever comes after command. */
if (c->prefixlist)
{
/* It is a prefix command; what comes after it is
a subcommand (e.g. "info "). */
list = complete_on_cmdlist (*c->prefixlist, p, word);
/* Insure that readline does the right thing
with respect to inserting quotes. */
rl_completer_word_break_characters =
gdb_completer_command_word_break_characters;
}
else if (c->enums)
{
list = complete_on_enum (c->enums, p, word);
rl_completer_word_break_characters =
gdb_completer_command_word_break_characters;
}
else
{
/* It is a normal command; what comes after it is
completed by the command's completer function. */
if (c->completer == filename_completer)
{
/* Many commands which want to complete on
file names accept several file names, as
in "run foo bar >>baz". So we don't want
to complete the entire text after the
command, just the last word. To this
end, we need to find the beginning of the
file name by starting at `word' and going
backwards. */
for (p = word;
p > tmp_command
&& strchr (gdb_completer_file_name_break_characters, p[-1]) == NULL;
p--)
;
rl_completer_word_break_characters =
gdb_completer_file_name_break_characters;
}
else if (c->completer == location_completer)
{
/* Commands which complete on locations want to
see the entire argument. */
for (p = word;
p > tmp_command
&& p[-1] != ' ' && p[-1] != '\t';
p--)
;
}
list = (*c->completer) (p, word);
}
}
else
{
/* The command is not followed by whitespace; we need to
complete on the command itself. e.g. "p" which is a
command itself but also can complete to "print", "ptype"
etc. */
char *q;
/* Find the command we are completing on. */
q = p;
while (q > tmp_command)
{
if (isalnum (q[-1]) || q[-1] == '-' || q[-1] == '_')
--q;
else
break;
}
list = complete_on_cmdlist (result_list, q, word);
/* Insure that readline does the right thing
with respect to inserting quotes. */
rl_completer_word_break_characters =
gdb_completer_command_word_break_characters;
}
}
else
{
/* There is non-whitespace beyond the command. */
if (c->prefixlist && !c->allow_unknown)
{
/* It is an unrecognized subcommand of a prefix command,
e.g. "info adsfkdj". */
list = NULL;
}
else if (c->enums)
{
list = complete_on_enum (c->enums, p, word);
}
else
{
/* It is a normal command. */
if (c->completer == filename_completer)
{
/* See the commentary above about the specifics
of file-name completion. */
for (p = word;
p > tmp_command
&& strchr (gdb_completer_file_name_break_characters, p[-1]) == NULL;
p--)
;
rl_completer_word_break_characters =
gdb_completer_file_name_break_characters;
}
else if (c->completer == location_completer)
{
for (p = word;
p > tmp_command
&& p[-1] != ' ' && p[-1] != '\t';
p--)
;
}
list = (*c->completer) (p, word);
}
}
}
return list;
}
/* Generate completions one by one for the completer. Each time we are
called return another potential completion to the caller.
line_completion just completes on commands or passes the buck to the
command's completer function, the stuff specific to symbol completion
is in make_symbol_completion_list.
TEXT is the caller's idea of the "word" we are looking at.
MATCHES is the number of matches that have currently been collected from
calling this completion function. When zero, then we need to initialize,
otherwise the initialization has already taken place and we can just
return the next potential completion string.
LINE_BUFFER is available to be looked at; it contains the entire text
of the line. POINT is the offset in that line of the cursor. You
should pretend that the line ends at POINT.
Returns NULL if there are no more completions, else a pointer to a string
which is a possible completion, it is the caller's responsibility to
free the string. */
char *
line_completion_function (char *text, int matches, char *line_buffer, int point)
{
static char **list = (char **) NULL; /* Cache of completions */
static int index; /* Next cached completion */
char *output = NULL;
if (matches == 0)
{
/* The caller is beginning to accumulate a new set of completions, so
we need to find all of them now, and cache them for returning one at
a time on future calls. */
if (list)
{
/* Free the storage used by LIST, but not by the strings inside.
This is because rl_complete_internal () frees the strings. */
xfree (list);
}
index = 0;
list = complete_line (text, line_buffer, point);
}
/* If we found a list of potential completions during initialization then
dole them out one at a time. The vector of completions is NULL
terminated, so after returning the last one, return NULL (and continue
to do so) each time we are called after that, until a new list is
available. */
if (list)
{
output = list[index];
if (output)
{
index++;
}
}
#if 0
/* Can't do this because readline hasn't yet checked the word breaks
for figuring out whether to insert a quote. */
if (output == NULL)
/* Make sure the word break characters are set back to normal for the
next time that readline tries to complete something. */
rl_completer_word_break_characters =
gdb_completer_word_break_characters;
#endif
return (output);
}