static int diff_flags = DIFF_INTERNAL | DIFF_FILLER | DIFF_CLOSE_OFF | DIFF_INLINE_CHAR;
static long diff_algorithm = XDF_INDENT_HEURISTIC;
+static int diff_word_gap = 5; // gap threshold for inline:word
# define LBUFLEN 50 // length of line in diff file
} while (pass++ < 4); // use limited number of passes to avoid excessive looping
}
+/*
+ * Refine inline word diff blocks by merging blocks that are only separated
+ * by whitespace or punctuation. This creates more coherent highlighting.
+ */
+ static void
+diff_refine_inline_word_highlight(diff_T *dp_orig, garray_T *linemap, int idx1,
+ linenr_T start_lnum)
+{
+ int pass = 1;
+ do
+ {
+ diff_T *dp = dp_orig;
+
+ while (dp != NULL && dp->df_next != NULL)
+ {
+ // Only merge blocks on the same line
+ if (dp->df_lnum[idx1] + dp->df_count[idx1] - 1 >= linemap[idx1]. ga_len
+ || dp->df_next->df_lnum[idx1] - 1 >= linemap[idx1]. ga_len)
+ {
+ dp = dp->df_next;
+ continue;
+ }
+
+ linemap_entry_T *entry1 =
+ &((linemap_entry_T *)linemap[idx1].ga_data)[dp->df_lnum[idx1] + dp->df_count[idx1] - 2];
+ linemap_entry_T *entry2 =
+ &((linemap_entry_T *)linemap[idx1].ga_data)[dp->df_next->df_lnum[idx1] - 1];
+
+ // Skip if blocks are on different lines
+ if (entry1->lineoff != entry2->lineoff)
+ {
+ dp = dp->df_next;
+ continue;
+ }
+
+ // Calculate the gap between blocks
+ int gap_start = entry1->byte_start + entry1->num_bytes;
+ int gap_end = entry2->byte_start;
+ int gap_size = gap_end - gap_start;
+
+ // Merge adjacent diff blocks separated by small gaps to reduce visual
+ // fragmentation. Gap threshold is set to 5 bytes which handles most
+ // common separators (spaces, punctuation, short variable names) while
+ // still preserving visually distinct changes.
+ if (gap_size <= 0 || gap_size > diff_word_gap)
+ {
+ dp = dp->df_next;
+ continue;
+ }
+
+ // Get the text between the two blocks
+ char_u *line = ml_get_buf(curtab->tp_diffbuf[idx1],
+ start_lnum + entry1->lineoff, FALSE);
+ char_u *gap_text = line + gap_start;
+
+ // Check if gap contains only whitespace and/or punctuation
+ bool only_non_word = true;
+ bool has_content = false;
+
+ for (int i = 0; i < gap_size && gap_text[i] != NUL; i++)
+ {
+ has_content = true;
+ int char_class = mb_get_class_buf(gap_text + i,
+ curtab->tp_diffbuf[idx1]);
+ // class 2 is word characters, if we find any, don't merge
+ if (char_class == 2)
+ {
+ only_non_word = false;
+ break;
+ }
+ }
+
+ // Merge if the gap is small and contains only non-word characters
+ if (has_content && only_non_word)
+ {
+ long total_change_bytes = 0;
+ for (int i = 0; i < DB_COUNT; i++)
+ {
+ if (curtab->tp_diffbuf[i] != NULL)
+ {
+ // count bytes in the first block
+ for (int k = 0; k < dp->df_count[i]; k++)
+ {
+ int idx = dp->df_lnum[i] + k - 1;
+ if (idx < linemap[i].ga_len)
+ total_change_bytes +=
+ ((linemap_entry_T *)linemap[i].ga_data)[idx].num_bytes;
+ }
+ // count bytes in the next block
+ for (int k = 0; k < dp->df_next->df_count[i]; k++)
+ {
+ int idx = dp->df_next->df_lnum[i] + k - 1;
+ if (idx < linemap[i].ga_len)
+ total_change_bytes +=
+ ((linemap_entry_T *)linemap[i].ga_data)[idx].num_bytes;
+ }
+ }
+ }
+
+ if (total_change_bytes >= gap_size * 2)
+ {
+ // Merge the blocks by extending the first block to include the next
+ for (int i = 0; i < DB_COUNT; i++)
+ {
+ if (curtab->tp_diffbuf[i] != NULL)
+ {
+ dp->df_count[i] = dp->df_next->df_lnum[i]
+ + dp->df_next->df_count[i] - dp->df_lnum[i];
+ }
+ }
+
+ diff_T *dp_next = dp->df_next;
+ dp->df_next = dp_next->df_next;
+ clear_diffblock(dp_next);
+
+ // Don't advance dp, check if can merge with the next block too
+ continue;
+ }
+ }
+
+ dp = dp->df_next;
+ }
+ } while (pass++ < 4); // use limited number of passes to avoid excessive looping
+}
+
/*
* Find the inline difference within a diff block among different buffers. Do
* this by splitting each block's content into characters or words, and then
}
diff_T *new_diff = curtab->tp_first_diff;
- if (diff_flags & DIFF_INLINE_CHAR && file1_idx != -1)
+ if (diff_flags & DIFF_INLINE_WORD && file1_idx != -1)
+ diff_refine_inline_word_highlight(new_diff, linemap, file1_idx, dp->df_lnum[file1_idx]);
+ else if (diff_flags & DIFF_INLINE_CHAR && file1_idx != -1)
diff_refine_inline_char_highlight(new_diff, linemap, file1_idx);
// After the diff, use the linemap to obtain the original line/col of the
projects / thirdparty / vim.git / commitdiff
