Per-path attribute based hunk header selection.
authorJunio C Hamano <gitster@pobox.com>
Fri, 6 Jul 2007 07:45:10 +0000 (00:45 -0700)
committerJunio C Hamano <gitster@pobox.com>
Fri, 6 Jul 2007 08:20:47 +0000 (01:20 -0700)
This makes"diff -p" hunk headers customizable via gitattributes mechanism.
It is based on Johannes's earlier patch that allowed to define a single
regexp to be used for everything.

The mechanism to arrive at the regexp that is used to define hunk header
is the same as other use of gitattributes.  You assign an attribute, funcname
(because "diff -p" typically uses the name of the function the patch is about
as the hunk header), a simple string value.  This can be one of the names of
built-in pattern (currently, "java" is defined) or a custom pattern name, to
be looked up from the configuration file.

  (in .gitattributes)
  *.java   funcname=java
  *.perl   funcname=perl

  (in .git/config)
  [funcname]
    java = ... # ugly and complicated regexp to override the built-in one.
    perl = ... # another ugly and complicated regexp to define a new one.

Signed-off-by: Junio C Hamano <gitster@pobox.com>
diff.c
diffcore.h
t/t4018-diff-funcname.sh [new file with mode: 0644]
xdiff-interface.c
xdiff-interface.h
xdiff/xdiff.h
xdiff/xemit.c

diff --git a/diff.c b/diff.c
index d10e848c792bba581d6bf3e30204a81204177983..04e7e91adfb83b7d25319d7307aec1584f9bdb71 100644 (file)
--- a/diff.c
+++ b/diff.c
@@ -1101,22 +1101,26 @@ static void emit_binary_diff(mmfile_t *one, mmfile_t *two)
 static void setup_diff_attr_check(struct git_attr_check *check)
 {
        static struct git_attr *attr_diff;
+       static struct git_attr *attr_diff_func_name;
 
        if (!attr_diff) {
                attr_diff = git_attr("diff", 4);
+               attr_diff_func_name = git_attr("funcname", 8);
        }
        check[0].attr = attr_diff;
+       check[1].attr = attr_diff_func_name;
 }
 
 static void diff_filespec_check_attr(struct diff_filespec *one)
 {
-       struct git_attr_check attr_diff_check[1];
+       struct git_attr_check attr_diff_check[2];
 
        if (one->checked_attr)
                return;
 
        setup_diff_attr_check(attr_diff_check);
        one->is_binary = 0;
+       one->hunk_header_ident = NULL;
 
        if (!git_checkattr(one->path, ARRAY_SIZE(attr_diff_check), attr_diff_check)) {
                const char *value;
@@ -1127,6 +1131,13 @@ static void diff_filespec_check_attr(struct diff_filespec *one)
                        ;
                else if (ATTR_FALSE(value))
                        one->is_binary = 1;
+
+               /* hunk header ident */
+               value = attr_diff_check[1].value;
+               if (ATTR_TRUE(value) || ATTR_FALSE(value) || ATTR_UNSET(value))
+                       ;
+               else
+                       one->hunk_header_ident = value;
        }
 
        if (!one->data && DIFF_FILE_VALID(one))
@@ -1143,6 +1154,82 @@ int diff_filespec_is_binary(struct diff_filespec *one)
        return one->is_binary;
 }
 
+static struct hunk_header_regexp {
+       char *name;
+       char *regexp;
+       struct hunk_header_regexp *next;
+} *hunk_header_regexp_list, **hunk_header_regexp_tail;
+
+static int hunk_header_config(const char *var, const char *value)
+{
+       static const char funcname[] = "funcname.";
+       struct hunk_header_regexp *hh;
+
+       if (prefixcmp(var, funcname))
+               return 0;
+       var += strlen(funcname);
+       for (hh = hunk_header_regexp_list; hh; hh = hh->next)
+               if (!strcmp(var, hh->name)) {
+                       free(hh->regexp);
+                       hh->regexp = xstrdup(value);
+                       return 0;
+               }
+       hh = xcalloc(1, sizeof(*hh));
+       hh->name = xstrdup(var);
+       hh->regexp = xstrdup(value);
+       hh->next = NULL;
+       *hunk_header_regexp_tail = hh;
+       return 0;
+}
+
+static const char *hunk_header_regexp(const char *ident)
+{
+       struct hunk_header_regexp *hh;
+
+       if (!hunk_header_regexp_tail) {
+               hunk_header_regexp_tail = &hunk_header_regexp_list;
+               git_config(hunk_header_config);
+       }
+       for (hh = hunk_header_regexp_list; hh; hh = hh->next)
+               if (!strcmp(ident, hh->name))
+                       return hh->regexp;
+       return NULL;
+}
+
+static const char *diff_hunk_header_regexp(struct diff_filespec *one)
+{
+       const char *ident, *regexp;
+
+       diff_filespec_check_attr(one);
+       ident = one->hunk_header_ident;
+
+       if (!ident)
+               /*
+                * If the config file has "funcname.default" defined, that
+                * regexp is used; otherwise NULL is returned and xemit uses
+                * the built-in default.
+                */
+               return hunk_header_regexp("default");
+
+       /* Look up custom "funcname.$ident" regexp from config. */
+       regexp = hunk_header_regexp(ident);
+       if (regexp)
+               return regexp;
+
+       /*
+        * And define built-in fallback patterns here.  Note that
+        * these can be overriden by the user's config settings.
+        */
+       if (!strcmp(ident, "java"))
+               return "!^[     ]*\\(catch\\|do\\|for\\|if\\|instanceof\\|"
+                       "new\\|return\\|switch\\|throw\\|while\\)\n"
+                       "^[     ]*\\(\\([       ]*"
+                       "[A-Za-z_][A-Za-z_0-9]*\\)\\{2,\\}"
+                       "[      ]*([^;]*$\\)";
+
+       return NULL;
+}
+
 static void builtin_diff(const char *name_a,
                         const char *name_b,
                         struct diff_filespec *one,
@@ -1217,6 +1304,11 @@ static void builtin_diff(const char *name_a,
                xdemitconf_t xecfg;
                xdemitcb_t ecb;
                struct emit_callback ecbdata;
+               const char *hunk_header_regexp;
+
+               hunk_header_regexp = diff_hunk_header_regexp(one);
+               if (!hunk_header_regexp)
+                       hunk_header_regexp = diff_hunk_header_regexp(two);
 
                memset(&xecfg, 0, sizeof(xecfg));
                memset(&ecbdata, 0, sizeof(ecbdata));
@@ -1226,6 +1318,8 @@ static void builtin_diff(const char *name_a,
                xpp.flags = XDF_NEED_MINIMAL | o->xdl_opts;
                xecfg.ctxlen = o->context;
                xecfg.flags = XDL_EMIT_FUNCNAMES;
+               if (hunk_header_regexp)
+                       xdiff_set_find_func(&xecfg, hunk_header_regexp);
                if (!diffopts)
                        ;
                else if (!prefixcmp(diffopts, "--unified="))
index dcab7e20bf6922a17f93ab268376d57765a8c6b8..05985147ea99ce54069163cd8c1372b7bb971c93 100644 (file)
@@ -27,6 +27,7 @@ struct diff_filespec {
        char *path;
        void *data;
        void *cnt_data;
+       const void *hunk_header_ident;
        unsigned long size;
        int xfrm_flags;          /* for use by the xfrm */
        unsigned short mode;     /* file mode */
diff --git a/t/t4018-diff-funcname.sh b/t/t4018-diff-funcname.sh
new file mode 100644 (file)
index 0000000..dc7a47b
--- /dev/null
@@ -0,0 +1,60 @@
+#!/bin/sh
+#
+# Copyright (c) 2007 Johannes E. Schindelin
+#
+
+test_description='Test custom diff function name patterns'
+
+. ./test-lib.sh
+
+LF='
+'
+
+cat > Beer.java << EOF
+public class Beer
+{
+       int special;
+       public static void main(String args[])
+       {
+               String s=" ";
+               for(int x = 99; x > 0; x--)
+               {
+                       System.out.print(x + " bottles of beer on the wall "
+                               + x + " bottles of beer\n"
+                               + "Take one down, pass it around, " + (x - 1)
+                               + " bottles of beer on the wall.\n");
+               }
+               System.out.print("Go to the store, buy some more,\n"
+                       + "99 bottles of beer on the wall.\n");
+       }
+}
+EOF
+
+sed 's/beer\\/beer,\\/' < Beer.java > Beer-correct.java
+
+test_expect_success 'default behaviour' '
+       git diff Beer.java Beer-correct.java |
+       grep "^@@.*@@ public class Beer"
+'
+
+test_expect_success 'preset java pattern' '
+       echo "*.java funcname=java" >.gitattributes &&
+       git diff Beer.java Beer-correct.java |
+       grep "^@@.*@@ public static void main("
+'
+
+git config funcname.java '!static
+!String
+[^     ].*s.*'
+
+test_expect_success 'custom pattern' '
+       git diff Beer.java Beer-correct.java |
+       grep "^@@.*@@ int special;$"
+'
+
+test_expect_success 'last regexp must not be negated' '
+       git config diff.functionnameregexp "!static" &&
+       ! git diff Beer.java Beer-correct.java
+'
+
+test_done
index e407cf11b1d8ba28be64eda7bcb505d9edb4d253..be866d12d38f6f1328f5fae8c7108176d4ecba70 100644 (file)
@@ -129,3 +129,74 @@ int buffer_is_binary(const char *ptr, unsigned long size)
                size = FIRST_FEW_BYTES;
        return !!memchr(ptr, 0, size);
 }
+
+struct ff_regs {
+       int nr;
+       struct ff_reg {
+               regex_t re;
+               int negate;
+       } *array;
+};
+
+static long ff_regexp(const char *line, long len,
+               char *buffer, long buffer_size, void *priv)
+{
+       char *line_buffer = xstrndup(line, len); /* make NUL terminated */
+       struct ff_regs *regs = priv;
+       regmatch_t pmatch[2];
+       int result = 0, i;
+
+       for (i = 0; i < regs->nr; i++) {
+               struct ff_reg *reg = regs->array + i;
+               if (reg->negate ^ !!regexec(&reg->re,
+                                       line_buffer, 2, pmatch, 0)) {
+                       free(line_buffer);
+                       return -1;
+               }
+       }
+       i = pmatch[1].rm_so >= 0 ? 1 : 0;
+       line += pmatch[i].rm_so;
+       result = pmatch[i].rm_eo - pmatch[i].rm_so;
+       if (result > buffer_size)
+               result = buffer_size;
+       else
+               while (result > 0 && (isspace(line[result - 1]) ||
+                                       line[result - 1] == '\n'))
+                       result--;
+       memcpy(buffer, line, result);
+       free(line_buffer);
+       return result;
+}
+
+void xdiff_set_find_func(xdemitconf_t *xecfg, const char *value)
+{
+       int i;
+       struct ff_regs *regs;
+
+       xecfg->find_func = ff_regexp;
+       regs = xecfg->find_func_priv = xmalloc(sizeof(struct ff_regs));
+       for (i = 0, regs->nr = 1; value[i]; i++)
+               if (value[i] == '\n')
+                       regs->nr++;
+       regs->array = xmalloc(regs->nr * sizeof(struct ff_reg));
+       for (i = 0; i < regs->nr; i++) {
+               struct ff_reg *reg = regs->array + i;
+               const char *ep = strchr(value, '\n'), *expression;
+               char *buffer = NULL;
+
+               reg->negate = (*value == '!');
+               if (reg->negate && i == regs->nr - 1)
+                       die("Last expression must not be negated: %s", value);
+               if (*value == '!')
+                       value++;
+               if (ep)
+                       expression = buffer = xstrndup(value, ep - value);
+               else
+                       expression = value;
+               if (regcomp(&reg->re, expression, 0))
+                       die("Invalid regexp to look for hunk header: %s", expression);
+               if (buffer)
+                       free(buffer);
+               value = ep + 1;
+       }
+}
index 536f4e4d9784e0e5ffdc108a5be4bf758d1578ba..fb742dbb6ba2125b970b9d9510e208ec99954a2d 100644 (file)
@@ -20,4 +20,6 @@ int parse_hunk_header(char *line, int len,
 int read_mmfile(mmfile_t *ptr, const char *filename);
 int buffer_is_binary(const char *ptr, unsigned long size);
 
+extern void xdiff_set_find_func(xdemitconf_t *xecfg, const char *line);
+
 #endif
index 9402bb0799978e181ba972a749a0ad849d2bd706..c00ddaa6e987407743d2c8877f9ca6e772f89c86 100644 (file)
@@ -73,9 +73,13 @@ typedef struct s_xdemitcb {
        int (*outf)(void *, mmbuffer_t *, int);
 } xdemitcb_t;
 
+typedef long (*find_func_t)(const char *line, long line_len, char *buffer, long buffer_size, void *priv);
+
 typedef struct s_xdemitconf {
        long ctxlen;
        unsigned long flags;
+       find_func_t find_func;
+       void *find_func_priv;
 } xdemitconf_t;
 
 typedef struct s_bdiffparam {
index 4b6e6391123ed92470f8cbaf8b87286e498dfe00..d3d9c845c6420e4881636d779c7029f900a0b067 100644 (file)
@@ -69,7 +69,24 @@ static xdchange_t *xdl_get_hunk(xdchange_t *xscr, xdemitconf_t const *xecfg) {
 }
 
 
-static void xdl_find_func(xdfile_t *xf, long i, char *buf, long sz, long *ll) {
+static long def_ff(const char *rec, long len, char *buf, long sz, void *priv)
+{
+       if (len > 0 &&
+                       (isalpha((unsigned char)*rec) || /* identifier? */
+                        *rec == '_' || /* also identifier? */
+                        *rec == '$')) { /* identifiers from VMS and other esoterico */
+               if (len > sz)
+                       len = sz;
+               while (0 < len && isspace((unsigned char)rec[len - 1]))
+                       len--;
+               memcpy(buf, rec, len);
+               return len;
+       }
+       return -1;
+}
+
+static void xdl_find_func(xdfile_t *xf, long i, char *buf, long sz, long *ll,
+               find_func_t ff, void *ff_priv) {
 
        /*
         * Be quite stupid about this for now.  Find a line in the old file
@@ -80,22 +97,12 @@ static void xdl_find_func(xdfile_t *xf, long i, char *buf, long sz, long *ll) {
        const char *rec;
        long len;
 
-       *ll = 0;
        while (i-- > 0) {
                len = xdl_get_rec(xf, i, &rec);
-               if (len > 0 &&
-                   (isalpha((unsigned char)*rec) || /* identifier? */
-                    *rec == '_' ||     /* also identifier? */
-                    *rec == '$')) {    /* mysterious GNU diff's invention */
-                       if (len > sz)
-                               len = sz;
-                       while (0 < len && isspace((unsigned char)rec[len - 1]))
-                               len--;
-                       memcpy(buf, rec, len);
-                       *ll = len;
+               if ((*ll = ff(rec, len, buf, sz, ff_priv)) >= 0)
                        return;
-               }
        }
+       *ll = 0;
 }
 
 
@@ -120,6 +127,7 @@ int xdl_emit_diff(xdfenv_t *xe, xdchange_t *xscr, xdemitcb_t *ecb,
        xdchange_t *xch, *xche;
        char funcbuf[80];
        long funclen = 0;
+       find_func_t ff = xecfg->find_func ?  xecfg->find_func : def_ff;
 
        if (xecfg->flags & XDL_EMIT_COMMON)
                return xdl_emit_common(xe, xscr, ecb, xecfg);
@@ -143,7 +151,8 @@ int xdl_emit_diff(xdfenv_t *xe, xdchange_t *xscr, xdemitcb_t *ecb,
 
                if (xecfg->flags & XDL_EMIT_FUNCNAMES) {
                        xdl_find_func(&xe->xdf1, s1, funcbuf,
-                                     sizeof(funcbuf), &funclen);
+                                     sizeof(funcbuf), &funclen,
+                                     ff, xecfg->find_func_priv);
                }
                if (xdl_emit_hunk_hdr(s1 + 1, e1 - s1, s2 + 1, e2 - s2,
                                      funcbuf, funclen, ecb) < 0)