forked from aniani/vim
updated for version 7.0100
This commit is contained in:
@@ -4682,7 +4682,8 @@ dozet:
|
||||
FIND_IDENT)) == 0)
|
||||
return;
|
||||
spell_add_word(ptr, len, nchar == 'w' || nchar == 'W',
|
||||
nchar == 'G' || nchar == 'W');
|
||||
(nchar == 'G' || nchar == 'W') ? 0
|
||||
: (int)cap->count1);
|
||||
}
|
||||
break;
|
||||
|
||||
|
240
src/spell.c
240
src/spell.c
@@ -378,7 +378,7 @@ typedef struct langp_S
|
||||
#define VIMSPELLMAGICL 10
|
||||
|
||||
/* file used for "zG" and "zW" */
|
||||
static char_u *temp_wordlist = NULL;
|
||||
static char_u *int_wordlist = NULL;
|
||||
|
||||
/*
|
||||
* Information used when looking for suggestions.
|
||||
@@ -588,6 +588,7 @@ static int spell_valid_case __ARGS((int origflags, int treeflags));
|
||||
static int no_spell_checking __ARGS((void));
|
||||
static void spell_load_lang __ARGS((char_u *lang));
|
||||
static char_u *spell_enc __ARGS((void));
|
||||
static void int_wordlist_spl __ARGS((char_u *fname));
|
||||
static void spell_load_cb __ARGS((char_u *fname, void *cookie));
|
||||
static slang_T *spell_load_file __ARGS((char_u *fname, char_u *lang, slang_T *old_lp, int silent));
|
||||
static char_u *read_cnt_string __ARGS((FILE *fd, int cnt_bytes, int *errp));
|
||||
@@ -693,21 +694,25 @@ static char *e_spell_trunc = N_("E758: Truncated spell file");
|
||||
* or when it's OK it remains unchanged.
|
||||
* This must only be called when 'spelllang' is not empty.
|
||||
*
|
||||
* "sug" is normally NULL. When looking for suggestions it points to
|
||||
* suginfo_T. It's passed as a void pointer to keep the struct local.
|
||||
* "capcol" is used to check for a Capitalised word after the end of a
|
||||
* sentence. If it's zero then perform the check. Return the column where to
|
||||
* check next, or -1 when no sentence end was found. If it's NULL then don't
|
||||
* worry.
|
||||
*
|
||||
* Returns the length of the word in bytes, also when it's OK, so that the
|
||||
* caller can skip over the word.
|
||||
*/
|
||||
int
|
||||
spell_check(wp, ptr, attrp)
|
||||
spell_check(wp, ptr, attrp, capcol)
|
||||
win_T *wp; /* current window */
|
||||
char_u *ptr;
|
||||
int *attrp;
|
||||
int *capcol; /* column to check for Capital */
|
||||
{
|
||||
matchinf_T mi; /* Most things are put in "mi" so that it can
|
||||
be passed to functions quickly. */
|
||||
int nrlen = 0; /* found a number first */
|
||||
int c;
|
||||
|
||||
/* A word never starts at a space or a control character. Return quickly
|
||||
* then, skipping over the character. */
|
||||
@@ -743,7 +748,24 @@ spell_check(wp, ptr, attrp)
|
||||
{
|
||||
mb_ptr_adv(mi.mi_fend);
|
||||
} while (*mi.mi_fend != NUL && spell_iswordp(mi.mi_fend, wp->w_buffer));
|
||||
|
||||
if (capcol != NULL && *capcol == 0 && wp->w_buffer->b_cap_prog != NULL)
|
||||
{
|
||||
/* Check word starting with capital letter. */
|
||||
#ifdef FEAT_MBYTE
|
||||
c = mb_ptr2char(ptr);
|
||||
#else
|
||||
c = *ptr;
|
||||
#endif
|
||||
if (!SPELL_ISUPPER(c))
|
||||
{
|
||||
*attrp = highlight_attr[HLF_SPC];
|
||||
return (int)(mi.mi_fend - ptr);
|
||||
}
|
||||
}
|
||||
}
|
||||
if (capcol != NULL)
|
||||
*capcol = -1;
|
||||
|
||||
/* We always use the characters up to the next non-word character,
|
||||
* also for bad words. */
|
||||
@@ -798,6 +820,17 @@ spell_check(wp, ptr, attrp)
|
||||
* skip over the character (try looking for a word after it). */
|
||||
else if (!SPELL_ISWORDP(ptr))
|
||||
{
|
||||
if (capcol != NULL && wp->w_buffer->b_cap_prog != NULL)
|
||||
{
|
||||
regmatch_T regmatch;
|
||||
|
||||
/* Check for end of sentence. */
|
||||
regmatch.regprog = wp->w_buffer->b_cap_prog;
|
||||
regmatch.rm_ic = FALSE;
|
||||
if (vim_regexec(®match, ptr, 0))
|
||||
*capcol = (int)(regmatch.endp[0] - ptr);
|
||||
}
|
||||
|
||||
#ifdef FEAT_MBYTE
|
||||
if (has_mbyte)
|
||||
return mb_ptr2len_check(ptr);
|
||||
@@ -1307,6 +1340,7 @@ spell_move_to(dir, allwords, curline)
|
||||
char_u *buf = NULL;
|
||||
int buflen = 0;
|
||||
int skip = 0;
|
||||
int capcol = -1;
|
||||
|
||||
if (no_spell_checking())
|
||||
return FAIL;
|
||||
@@ -1339,6 +1373,14 @@ spell_move_to(dir, allwords, curline)
|
||||
break;
|
||||
}
|
||||
|
||||
/* In first line check first word for Capital. */
|
||||
if (lnum == 1)
|
||||
capcol = 0;
|
||||
|
||||
/* For checking first word with a capital skip white space. */
|
||||
if (capcol == 0)
|
||||
capcol = skipwhite(line) - line;
|
||||
|
||||
/* Copy the line into "buf" and append the start of the next line if
|
||||
* possible. */
|
||||
STRCPY(buf, line);
|
||||
@@ -1357,7 +1399,7 @@ spell_move_to(dir, allwords, curline)
|
||||
|
||||
/* start of word */
|
||||
attr = 0;
|
||||
len = spell_check(curwin, p, &attr);
|
||||
len = spell_check(curwin, p, &attr, &capcol);
|
||||
|
||||
if (attr != 0)
|
||||
{
|
||||
@@ -1403,6 +1445,7 @@ spell_move_to(dir, allwords, curline)
|
||||
|
||||
/* advance to character after the word */
|
||||
p += len;
|
||||
capcol -= len;
|
||||
}
|
||||
|
||||
if (curline)
|
||||
@@ -1421,6 +1464,7 @@ spell_move_to(dir, allwords, curline)
|
||||
if (lnum == 1)
|
||||
break;
|
||||
--lnum;
|
||||
capcol = -1;
|
||||
}
|
||||
else
|
||||
{
|
||||
@@ -1434,6 +1478,13 @@ spell_move_to(dir, allwords, curline)
|
||||
skip = p - endp;
|
||||
else
|
||||
skip = 0;
|
||||
|
||||
/* Capscol skips over the inserted space. */
|
||||
--capcol;
|
||||
|
||||
/* But after empty line check first word in next line */
|
||||
if (*skipwhite(line) == NUL)
|
||||
capcol = 0;
|
||||
}
|
||||
|
||||
line_breakcheck();
|
||||
@@ -1528,6 +1579,18 @@ spell_enc()
|
||||
return (char_u *)"latin1";
|
||||
}
|
||||
|
||||
/*
|
||||
* Get the name of the .spl file for the internal wordlist into
|
||||
* "fname[MAXPATHL]".
|
||||
*/
|
||||
static void
|
||||
int_wordlist_spl(fname)
|
||||
char_u *fname;
|
||||
{
|
||||
vim_snprintf((char *)fname, MAXPATHL, "%s.%s.spl",
|
||||
int_wordlist, spell_enc());
|
||||
}
|
||||
|
||||
/*
|
||||
* Allocate a new slang_T.
|
||||
* Caller must fill "sl_next".
|
||||
@@ -2445,24 +2508,14 @@ did_set_spelllang(buf)
|
||||
int c;
|
||||
char_u lang[MAXWLEN + 1];
|
||||
char_u spf_name[MAXPATHL];
|
||||
int load_spf;
|
||||
int len;
|
||||
char_u *p;
|
||||
int round;
|
||||
char_u *spf;
|
||||
|
||||
ga_init2(&ga, sizeof(langp_T), 2);
|
||||
clear_midword(buf);
|
||||
|
||||
/* Make the name of the .spl file associated with 'spellfile'. */
|
||||
if (*buf->b_p_spf == NUL)
|
||||
load_spf = FALSE;
|
||||
else
|
||||
{
|
||||
vim_snprintf((char *)spf_name, sizeof(spf_name), "%s.spl",
|
||||
buf->b_p_spf);
|
||||
load_spf = TRUE;
|
||||
}
|
||||
|
||||
/* loop over comma separated language names. */
|
||||
for (splp = buf->b_p_spl; *splp != NUL; )
|
||||
{
|
||||
@@ -2541,32 +2594,37 @@ did_set_spelllang(buf)
|
||||
LANGP_ENTRY(ga, ga.ga_len)->lp_region = region_mask;
|
||||
++ga.ga_len;
|
||||
use_midword(lp, buf);
|
||||
|
||||
/* Check if this is the spell file related to 'spellfile'. */
|
||||
if (load_spf && fullpathcmp(spf_name, lp->sl_fname, FALSE)
|
||||
== FPC_SAME)
|
||||
load_spf = FALSE;
|
||||
}
|
||||
}
|
||||
|
||||
/* round 1: load 'spellfile', if needed.
|
||||
* round 2: load temp_wordlist, if possible. */
|
||||
for (round = 1; round <= 2; ++round)
|
||||
/* round 0: load int_wordlist, if possible.
|
||||
* round 1: load first name in 'spellfile'.
|
||||
* round 2: load second name in 'spellfile.
|
||||
* etc. */
|
||||
spf = curbuf->b_p_spf;
|
||||
for (round = 0; round == 0 || *spf != NUL; ++round)
|
||||
{
|
||||
if (round == 1)
|
||||
if (round == 0)
|
||||
{
|
||||
/* Make sure the 'spellfile' file is loaded. It may be in
|
||||
* 'runtimepath', then it's probably loaded above already.
|
||||
* Otherwise load it here. */
|
||||
if (!load_spf)
|
||||
/* Internal wordlist, if there is one. */
|
||||
if (int_wordlist == NULL)
|
||||
continue;
|
||||
int_wordlist_spl(spf_name);
|
||||
}
|
||||
else
|
||||
{
|
||||
if (temp_wordlist == NULL)
|
||||
/* One entry in 'spellfile'. */
|
||||
copy_option_part(&spf, spf_name, MAXPATHL - 5, ",");
|
||||
STRCAT(spf_name, ".spl");
|
||||
|
||||
/* If it was already found above then skip it. */
|
||||
for (c = 0; c < ga.ga_len; ++c)
|
||||
if (fullpathcmp(spf_name,
|
||||
LANGP_ENTRY(ga, c)->lp_slang->sl_fname,
|
||||
FALSE) == FPC_SAME)
|
||||
break;
|
||||
if (c < ga.ga_len)
|
||||
continue;
|
||||
vim_snprintf((char *)spf_name, sizeof(spf_name), "%s.%s.spl",
|
||||
temp_wordlist, spell_enc());
|
||||
}
|
||||
|
||||
/* Check if it was loaded already. */
|
||||
@@ -2576,17 +2634,17 @@ did_set_spelllang(buf)
|
||||
if (lp == NULL)
|
||||
{
|
||||
/* Not loaded, try loading it now. The language name includes the
|
||||
* region name, the region is ignored otherwise. for
|
||||
* temp_wordlist use an arbitrary name. */
|
||||
if (round == 1)
|
||||
* region name, the region is ignored otherwise. for int_wordlist
|
||||
* use an arbitrary name. */
|
||||
if (round == 0)
|
||||
STRCPY(lang, "internal wordlist");
|
||||
else
|
||||
{
|
||||
vim_strncpy(lang, gettail(buf->b_p_spf), MAXWLEN);
|
||||
vim_strncpy(lang, gettail(spf_name), MAXWLEN);
|
||||
p = vim_strchr(lang, '.');
|
||||
if (p != NULL)
|
||||
*p = NUL; /* truncate at ".encoding.add" */
|
||||
}
|
||||
else
|
||||
STRCPY(lang, "temp wordlist");
|
||||
lp = spell_load_file(spf_name, lang, NULL, TRUE);
|
||||
}
|
||||
if (lp != NULL && ga_grow(&ga, 1) == OK)
|
||||
@@ -2765,6 +2823,7 @@ spell_free_all()
|
||||
{
|
||||
slang_T *lp;
|
||||
buf_T *buf;
|
||||
char_u fname[MAXPATHL];
|
||||
|
||||
/* Go through all buffers and handle 'spelllang'. */
|
||||
for (buf = firstbuf; buf != NULL; buf = buf->b_next)
|
||||
@@ -2777,11 +2836,14 @@ spell_free_all()
|
||||
slang_free(lp);
|
||||
}
|
||||
|
||||
if (temp_wordlist != NULL)
|
||||
if (int_wordlist != NULL)
|
||||
{
|
||||
mch_remove(temp_wordlist);
|
||||
vim_free(temp_wordlist);
|
||||
temp_wordlist = NULL;
|
||||
/* Delete the internal wordlist and its .spl file */
|
||||
mch_remove(int_wordlist);
|
||||
int_wordlist_spl(fname);
|
||||
mch_remove(fname);
|
||||
vim_free(int_wordlist);
|
||||
int_wordlist = NULL;
|
||||
}
|
||||
|
||||
init_spell_chartab();
|
||||
@@ -5172,42 +5234,48 @@ mkspell(fcount, fnames, ascii, overwrite, added_word)
|
||||
|
||||
|
||||
/*
|
||||
* ":spellgood {word}"
|
||||
* ":spellwrong {word}"
|
||||
* ":[count]spellgood {word}"
|
||||
* ":[count]spellwrong {word}"
|
||||
*/
|
||||
void
|
||||
ex_spell(eap)
|
||||
exarg_T *eap;
|
||||
{
|
||||
spell_add_word(eap->arg, STRLEN(eap->arg), eap->cmdidx == CMD_spellwrong,
|
||||
eap->forceit);
|
||||
eap->forceit ? 0 : (int)eap->line2);
|
||||
}
|
||||
|
||||
/*
|
||||
* Add "word[len]" to 'spellfile' as a good or bad word.
|
||||
*/
|
||||
void
|
||||
spell_add_word(word, len, bad, temp)
|
||||
spell_add_word(word, len, bad, index)
|
||||
char_u *word;
|
||||
int len;
|
||||
int bad;
|
||||
int temp; /* "zG" and "zW": use temp word list */
|
||||
int index; /* "zG" and "zW": zero, otherwise index in
|
||||
'spellfile' */
|
||||
{
|
||||
FILE *fd;
|
||||
buf_T *buf;
|
||||
buf_T *buf = NULL;
|
||||
int new_spf = FALSE;
|
||||
struct stat st;
|
||||
char_u *fname;
|
||||
char_u fnamebuf[MAXPATHL];
|
||||
char_u line[MAXWLEN * 2];
|
||||
long fpos, fpos_next = 0;
|
||||
int i;
|
||||
char_u *spf;
|
||||
|
||||
if (temp) /* use temp word list */
|
||||
if (index == 0) /* use internal wordlist */
|
||||
{
|
||||
if (temp_wordlist == NULL)
|
||||
if (int_wordlist == NULL)
|
||||
{
|
||||
temp_wordlist = vim_tempname('s');
|
||||
if (temp_wordlist == NULL)
|
||||
int_wordlist = vim_tempname('s');
|
||||
if (int_wordlist == NULL)
|
||||
return;
|
||||
}
|
||||
fname = temp_wordlist;
|
||||
fname = int_wordlist;
|
||||
}
|
||||
else
|
||||
{
|
||||
@@ -5224,8 +5292,20 @@ spell_add_word(word, len, bad, temp)
|
||||
return;
|
||||
}
|
||||
|
||||
for (spf = curbuf->b_p_spf, i = 1; *spf != NUL; ++i)
|
||||
{
|
||||
copy_option_part(&spf, fnamebuf, MAXPATHL, ",");
|
||||
if (i == index)
|
||||
break;
|
||||
if (*spf == NUL)
|
||||
{
|
||||
EMSGN(_("E765: 'spellfile' does not have %ld enties"), index);
|
||||
return;
|
||||
}
|
||||
}
|
||||
|
||||
/* Check that the user isn't editing the .add file somewhere. */
|
||||
buf = buflist_findname_exp(curbuf->b_p_spf);
|
||||
buf = buflist_findname_exp(fnamebuf);
|
||||
if (buf != NULL && buf->b_ml.ml_mfp == NULL)
|
||||
buf = NULL;
|
||||
if (buf != NULL && bufIsChanged(buf))
|
||||
@@ -5234,7 +5314,37 @@ spell_add_word(word, len, bad, temp)
|
||||
return;
|
||||
}
|
||||
|
||||
fname = curbuf->b_p_spf;
|
||||
fname = fnamebuf;
|
||||
}
|
||||
|
||||
if (bad)
|
||||
{
|
||||
/* When the word also appears as good word we need to remove that one,
|
||||
* since its flags sort before the one with WF_BANNED. */
|
||||
fd = mch_fopen((char *)fname, "r");
|
||||
if (fd != NULL)
|
||||
{
|
||||
while (!vim_fgets(line, MAXWLEN * 2, fd))
|
||||
{
|
||||
fpos = fpos_next;
|
||||
fpos_next = ftell(fd);
|
||||
if (STRNCMP(word, line, len) == 0
|
||||
&& (line[len] == '/' || line[len] < ' '))
|
||||
{
|
||||
/* Found duplicate word. Remove it by writing a '#' at
|
||||
* the start of the line. Mixing reading and writing
|
||||
* doesn't work for all systems, close the file first. */
|
||||
fclose(fd);
|
||||
fd = mch_fopen((char *)fname, "r+");
|
||||
if (fd == NULL)
|
||||
break;
|
||||
if (fseek(fd, fpos, SEEK_SET) == 0)
|
||||
fputc('#', fd);
|
||||
fseek(fd, fpos_next, SEEK_SET);
|
||||
}
|
||||
}
|
||||
fclose(fd);
|
||||
}
|
||||
}
|
||||
|
||||
fd = mch_fopen((char *)fname, "a");
|
||||
@@ -6088,7 +6198,7 @@ spell_find_suggest(badptr, su, maxcount, banbadword)
|
||||
int maxcount;
|
||||
int banbadword; /* don't include badword in suggestions */
|
||||
{
|
||||
int attr;
|
||||
int attr = 0;
|
||||
char_u buf[MAXPATHL];
|
||||
char_u *p;
|
||||
int do_combine = FALSE;
|
||||
@@ -6096,6 +6206,7 @@ spell_find_suggest(badptr, su, maxcount, banbadword)
|
||||
#ifdef FEAT_EVAL
|
||||
static int expr_busy = FALSE;
|
||||
#endif
|
||||
int c;
|
||||
|
||||
/*
|
||||
* Set the info in "*su".
|
||||
@@ -6108,7 +6219,7 @@ spell_find_suggest(badptr, su, maxcount, banbadword)
|
||||
hash_init(&su->su_banned);
|
||||
|
||||
su->su_badptr = badptr;
|
||||
su->su_badlen = spell_check(curwin, su->su_badptr, &attr);
|
||||
su->su_badlen = spell_check(curwin, su->su_badptr, &attr, NULL);
|
||||
su->su_maxcount = maxcount;
|
||||
su->su_maxscore = SCORE_MAXINIT;
|
||||
|
||||
@@ -6120,6 +6231,21 @@ spell_find_suggest(badptr, su, maxcount, banbadword)
|
||||
/* get caps flags for bad word */
|
||||
su->su_badflags = captype(su->su_badptr, su->su_badptr + su->su_badlen);
|
||||
|
||||
/* If the word is not capitalised and spell_check() doesn't consider the
|
||||
* word to be bad then it might need to be capitalised. Add a suggestion
|
||||
* for that. */
|
||||
#ifdef FEAT_MBYTE
|
||||
c = mb_ptr2char(su->su_badptr);
|
||||
#else
|
||||
c = *p;
|
||||
#endif
|
||||
if (!SPELL_ISUPPER(c) && attr == 0)
|
||||
{
|
||||
make_case_word(su->su_badword, buf, WF_ONECAP);
|
||||
add_suggestion(su, &su->su_ga, buf, su->su_badlen, SCORE_ICASE,
|
||||
0, TRUE);
|
||||
}
|
||||
|
||||
/* Ban the bad word itself. It may appear in another region. */
|
||||
if (banbadword)
|
||||
add_banned(su, su->su_badword);
|
||||
|
@@ -1137,6 +1137,7 @@ enum hlf_value
|
||||
, HLF_TXD /* Text Changed in diff line */
|
||||
, HLF_SC /* Sign column */
|
||||
, HLF_SPB /* SpellBad */
|
||||
, HLF_SPC /* SpellCap */
|
||||
, HLF_SPR /* SpellRare */
|
||||
, HLF_SPL /* SpellLocal */
|
||||
, HLF_COUNT /* MUST be the last one */
|
||||
@@ -1145,7 +1146,7 @@ enum hlf_value
|
||||
/* the HL_FLAGS must be in the same order as the HLF_ enums! */
|
||||
#define HL_FLAGS {'8', '@', 'd', 'e', 'h', 'i', 'l', 'm', 'M', \
|
||||
'n', 'r', 's', 'S', 'c', 't', 'v', 'V', 'w', 'W', \
|
||||
'f', 'F', 'A', 'C', 'D', 'T', '>', 'B', 'R', 'L'}
|
||||
'f', 'F', 'A', 'C', 'D', 'T', '>', 'B', 'P', 'R', 'L'}
|
||||
|
||||
/*
|
||||
* Boolean constants
|
||||
|
Reference in New Issue
Block a user