diff --git a/cpp/parser/GumboNormalizer.c b/cpp/parser/GumboNormalizer.c index c9bea6b27..e2311a6f0 100644 --- a/cpp/parser/GumboNormalizer.c +++ b/cpp/parser/GumboNormalizer.c @@ -411,6 +411,53 @@ static void emit_one_attr(buffer_t *out, GumboElement *el, } } +/* Tags that may carry a text-align style in our canonical output */ +static bool is_alignable_tag(const char *tag_name) { + return strcmp(tag_name, "p") == 0 || strcmp(tag_name, "ul") == 0 || + strcmp(tag_name, "ol") == 0 || strcmp(tag_name, "h1") == 0 || + strcmp(tag_name, "h2") == 0 || strcmp(tag_name, "h3") == 0 || + strcmp(tag_name, "h4") == 0 || strcmp(tag_name, "h5") == 0 || + strcmp(tag_name, "h6") == 0; +} + +static const char *canonical_alignment(const char *val, size_t val_len) { + static const char *aligns[] = {"left", "center", "right", "justify"}; + for (size_t i = 0; i < 4; i++) { + size_t alen = strlen(aligns[i]); + if (alen != val_len) + continue; + bool match = true; + for (size_t j = 0; j < val_len; j++) { + if (tolower((unsigned char)val[j]) != aligns[i][j]) { + match = false; + break; + } + } + if (match) + return aligns[i]; + } + return NULL; +} + +static void emit_alignment(GumboElement *el, const char *tag_name, + buffer_t *out) { + if (!is_alignable_tag(tag_name)) + return; + const char *style = get_attr(el, "style"); + if (!style) + return; + size_t vlen; + const char *val = find_css_value(style, strlen(style), "text-align", &vlen); + if (!val) + return; + const char *canon = canonical_alignment(val, vlen); + if (!canon) + return; + buffer_append_str(out, " style=\"text-align: "); + buffer_append_str(out, canon); + buffer_append_str(out, "\""); +} + static bool is_checkbox_list(GumboElement *el) { const char *val = get_attr(el, "data-type"); if (val && (strcmp(val, "checkbox") == 0 || strcmp(val, "checkboxList") == 0)) { @@ -452,6 +499,7 @@ static void emit_attributes(GumboElement *el, const char *tag_name, if (is_checkbox_list(el)) { buffer_append_str(out, " data-type=\"checkbox\""); } + emit_alignment(el, tag_name, out); } else if (strcmp(tag_name, "li") == 0) { const char *data_checked = get_attr(el, "data-checked"); const char *aria_checked = get_attr(el, "aria-checked"); @@ -468,6 +516,9 @@ static void emit_attributes(GumboElement *el, const char *tag_name, emit_one_attr(out, el, "id"); emit_one_attr(out, el, "text"); emit_one_attr(out, el, "indicator"); + } else { + /* preserve text-align */ + emit_alignment(el, tag_name, out); } } @@ -497,9 +548,13 @@ static void walk_node(GumboNode *node, buffer_t *out); static void flatten_bq_node(GumboNode *node, buffer_t *ib, buffer_t *out); -static void flush_inline_p(buffer_t *ib, buffer_t *out) { +static void flush_inline_p(buffer_t *ib, buffer_t *out, + GumboElement *align_el) { if (ib->len > 0) { - buffer_append_str(out, "

"); + buffer_append_str(out, ""); buffer_append(out, ib->data, ib->len); buffer_append_str(out, "

"); buffer_clear(ib); @@ -526,13 +581,14 @@ static void flatten_bq_node(GumboNode *node, buffer_t *ib, buffer_t *out) { return; } if (is_br_node(node)) { - flush_inline_p(ib, out); + flush_inline_p(ib, out, NULL); return; } if (is_block_producing(node) || is_blockquote_node(node)) { - flush_inline_p(ib, out); + flush_inline_p(ib, out, NULL); flatten_bq_children(node, ib, out); - flush_inline_p(ib, out); + // The flattened block becomes a

; carry over its text-align (if any). + flush_inline_p(ib, out, &node->v.element); return; } walk_node(node, ib); @@ -660,7 +716,7 @@ static void walk_children(GumboNode *node, buffer_t *out) { flatten_bq_children(children->data[i], &bq_ib, out); i++; } - flush_inline_p(&bq_ib, out); + flush_inline_p(&bq_ib, out, NULL); free(bq_ib.data); buffer_append_str(out, ""); continue; @@ -675,7 +731,7 @@ static void walk_children(GumboNode *node, buffer_t *out) { child = children->data[i]; if (is_br_node(child)) { if (ib.len > 0) - flush_inline_p(&ib, out); + flush_inline_p(&ib, out, NULL); else buffer_append_str(out, "
"); i++; @@ -683,7 +739,7 @@ static void walk_children(GumboNode *node, buffer_t *out) { } /* Transparent inline wrapper for block/bq children */ if (is_element(child) && has_block_or_bq_child(child)) { - flush_inline_p(&ib, out); + flush_inline_p(&ib, out, NULL); walk_children(child, out); i++; continue; @@ -691,7 +747,7 @@ static void walk_children(GumboNode *node, buffer_t *out) { walk_node(child, &ib); i++; } - flush_inline_p(&ib, out); + flush_inline_p(&ib, out, NULL); free(ib.data); continue; } diff --git a/cpp/tests/GumboParserTest.cpp b/cpp/tests/GumboParserTest.cpp index 124e44642..0f5586969 100644 --- a/cpp/tests/GumboParserTest.cpp +++ b/cpp/tests/GumboParserTest.cpp @@ -509,3 +509,65 @@ TEST(GumboParserTest, BrRemappings) { "

Asdasdasd



Sent with Net

"); } + +// Preserve text alignment +TEST(GumboParserTest, TextAlignment) { + EXPECT_EQ(GumboParser::normalizeHtml("

x

"), + "

x

"); + EXPECT_EQ(GumboParser::normalizeHtml("

x

"), + "

x

"); + EXPECT_EQ(GumboParser::normalizeHtml("

x

"), + "

x

"); + EXPECT_EQ( + GumboParser::normalizeHtml("

x

"), + "

x

"); + + EXPECT_EQ(GumboParser::normalizeHtml( + ""), + ""); + EXPECT_EQ(GumboParser::normalizeHtml( + "
  1. x
"), + "
  1. x
"); + EXPECT_EQ(GumboParser::normalizeHtml( + ""), + ""); + + EXPECT_EQ( + GumboParser::normalizeHtml("

x

"), + "

x

"); + EXPECT_EQ( + GumboParser::normalizeHtml("
x
"), + "
x
"); + + // Value is normalized to lowercase + EXPECT_EQ(GumboParser::normalizeHtml("

x

"), + "

x

"); + + // Coexists with inline formatting on the same tag + EXPECT_EQ(GumboParser::normalizeHtml( + "

x

"), + "

x

"); + + // Invalid value is stripped + EXPECT_EQ(GumboParser::normalizeHtml("

x

"), + "

x

"); + + // Not emitted on non-alignable tags + EXPECT_EQ(GumboParser::normalizeHtml( + ""), + ""); + EXPECT_EQ(GumboParser::normalizeHtml( + "
x
"), + "

x

"); + + // Preserved per-paragraph when a

blocks are flattened + EXPECT_EQ(GumboParser::normalizeHtml( + "

l

" + "

c

" + "

r

"), + "

l

" + "

c

" + "

r

"); +} diff --git a/src/web/__tests__/htmlNormalizer.test.ts b/src/web/__tests__/htmlNormalizer.test.ts index ac5dba46c..e2ebcf3ed 100644 --- a/src/web/__tests__/htmlNormalizer.test.ts +++ b/src/web/__tests__/htmlNormalizer.test.ts @@ -497,4 +497,82 @@ describe('htmlNormalizer', () => { expect(normalizeHtml(input)).toBe(expected); }); }); + + // Preserve text alignment + describe('TextAlignment', () => { + test.each([ + [ + '

x

', + '

x

', + ], + [ + '

x

', + '

x

', + ], + [ + '

x

', + '

x

', + ], + [ + '

x

', + '

x

', + ], + + [ + '', + '', + ], + [ + '
  1. x
', + '
  1. x
', + ], + [ + '', + '', + ], + + [ + '

x

', + '

x

', + ], + [ + '
x
', + '
x
', + ], + + // Value is normalized to lowercase + [ + '

x

', + '

x

', + ], + + // Coexists with inline formatting on the same tag + [ + '

x

', + '

x

', + ], + + // Invalid value is stripped + ['

x

', '

x

'], + + // Not emitted on non-alignable tags + ['', ''], + [ + '
x
', + '

x

', + ], + + // Preserved per-paragraph when a

blocks are flattened + [ + '

l

' + + '

c

' + + '

r

', + '

l

' + + '

c

' + + '

r

', + ], + ])('%s → %s', (input, expected) => { + expect(normalizeHtml(input)).toBe(expected); + }); + }); }); diff --git a/src/web/normalization/htmlNormalizer.ts b/src/web/normalization/htmlNormalizer.ts index 670f40959..76237885a 100644 --- a/src/web/normalization/htmlNormalizer.ts +++ b/src/web/normalization/htmlNormalizer.ts @@ -231,6 +231,29 @@ function emitStylesClose(s: CssStyles): string { return out; } +// Tags that may carry a text-align style in our canonical output +const ALIGN_TAGS = new Set([ + 'p', + 'ul', + 'ol', + 'h1', + 'h2', + 'h3', + 'h4', + 'h5', + 'h6', +]); +const ALIGN_VALUES = new Set(['left', 'center', 'right', 'justify']); + +function emitAlignment(el: Element, name: string): string { + if (!ALIGN_TAGS.has(name)) return ''; + const align = findCssValue(el.getAttribute('style') ?? '', 'text-align'); + if (!align) return ''; + const lower = align.toLowerCase(); + if (!ALIGN_VALUES.has(lower)) return ''; + return ` style="text-align: ${lower}"`; +} + function emitOneAttr(el: Element, attr: string): string { const val = el.getAttribute(attr); if (val == null || val === '') return ''; @@ -250,7 +273,10 @@ function emitAttributes(el: Element, name: string): string { emitOneAttr(el, 'height') ); case 'ul': - return isCheckboxList(el) ? ' data-type="checkbox"' : ''; + return ( + (isCheckboxList(el) ? ' data-type="checkbox"' : '') + + emitAlignment(el, name) + ); case 'li': // "" is U+F0FE (MS Word checked box); often encoded as "\xEF\x83\xBE" in UTF-8. const isChecked = @@ -266,7 +292,8 @@ function emitAttributes(el: Element, name: string): string { emitOneAttr(el, 'indicator') ); default: - return ''; + // preserve text-align + return emitAlignment(el, name); } } @@ -323,9 +350,13 @@ function escapeText(s: string): string { // --- Blockquote content flattening --- -function flushInlineP(ib: { buf: string }, out: { buf: string }): void { +function flushInlineP( + ib: { buf: string }, + out: { buf: string }, + attrs = '' +): void { if (ib.buf.length > 0) { - out.buf += `

${ib.buf}

`; + out.buf += `${ib.buf}

`; ib.buf = ''; } } @@ -358,7 +389,8 @@ function flattenBqNode( if (isBlockProducing(node) || isBlockquoteNode(node)) { flushInlineP(ib, out); flattenBqChildren(node, ib, out); - flushInlineP(ib, out); + // The flattened block becomes a

; carry over its text-align (if any). + flushInlineP(ib, out, emitAlignment(node, 'p')); return; } walkNode(node, ib);