Remove lots of string literals Using strings in helper functions causes a lot of string-to-[]byte allocations. This fix is centered around converging the tag() helper func to the []byte lingo. In order to do that, a lot of string literals have moved to global variables, where string to []byte conversion can happen once.
Vytautas Šaltenis vytas@rtfb.lt
Sat, 10 Sep 2016 12:39:22 +0300
1 files changed,
163 insertions(+),
58 deletions(-)
jump to
M
html.go
→
html.go
@@ -315,15 +315,19 @@ }
return attrs } -func tag(name string, attrs []string, selfClosing bool) []byte { - result := "<" + name - if attrs != nil && len(attrs) > 0 { - result += " " + strings.Join(attrs, " ") - } - if selfClosing { - result += " /" +var ( + gtBytes = []byte{'>'} + spaceBytes = []byte{' '} +) + +func (r *HTMLRenderer) tag(w io.Writer, name []byte, attrs []string) { + w.Write(name) + if len(attrs) > 0 { + w.Write(spaceBytes) + w.Write([]byte(strings.Join(attrs, " "))) } - return []byte(result + ">") + w.Write(gtBytes) + r.lastOutputLen = 1 } func footnoteRef(prefix string, node *Node) []byte {@@ -397,6 +401,95 @@ r.out(w, []byte{'\n'})
} } +var ( + brTag = []byte("<br>") + brXHTMLTag = []byte("<br />") + emTag = []byte("<em>") + emCloseTag = []byte("</em>") + strongTag = []byte("<strong>") + strongCloseTag = []byte("</strong>") + delTag = []byte("<del>") + delCloseTag = []byte("</del>") + ttTag = []byte("<tt>") + ttCloseTag = []byte("</tt>") + aTag = []byte("<a") + aCloseTag = []byte("</a>") + preTag = []byte("<pre>") + preCloseTag = []byte("</pre>") + codeTag = []byte("<code>") + codeCloseTag = []byte("</code>") + pTag = []byte("<p>") + pCloseTag = []byte("</p>") + blockquoteTag = []byte("<blockquote>") + blockquoteCloseTag = []byte("</blockquote>") + hrTag = []byte("<hr>") + hrXHTMLTag = []byte("<hr />") + ulTag = []byte("<ul>") + ulCloseTag = []byte("</ul>") + olTag = []byte("<ol>") + olCloseTag = []byte("</ol>") + dlTag = []byte("<dl>") + dlCloseTag = []byte("</dl>") + liTag = []byte("<li>") + liCloseTag = []byte("</li>") + ddTag = []byte("<dd>") + ddCloseTag = []byte("</dd>") + dtTag = []byte("<dt>") + dtCloseTag = []byte("</dt>") + tableTag = []byte("<table>") + tableCloseTag = []byte("</table>") + tdTag = []byte("<td") + tdCloseTag = []byte("</td>") + thTag = []byte("<th") + thCloseTag = []byte("</th>") + theadTag = []byte("<thead>") + theadCloseTag = []byte("</thead>") + tbodyTag = []byte("<tbody>") + tbodyCloseTag = []byte("</tbody>") + trTag = []byte("<tr>") + trCloseTag = []byte("</tr>") + h1Tag = []byte("<h1") + h1CloseTag = []byte("</h1>") + h2Tag = []byte("<h2") + h2CloseTag = []byte("</h2>") + h3Tag = []byte("<h3") + h3CloseTag = []byte("</h3>") + h4Tag = []byte("<h4") + h4CloseTag = []byte("</h4>") + h5Tag = []byte("<h5") + h5CloseTag = []byte("</h5>") + h6Tag = []byte("<h6") + h6CloseTag = []byte("</h6>") + + footnotesDivBytes = []byte("\n<div class=\"footnotes\">\n\n") + footnotesCloseDivBytes = []byte("\n</div>\n") +) + +func headerTagsFromLevel(level int) ([]byte, []byte) { + switch level { + case 1: + return h1Tag, h1CloseTag + case 2: + return h2Tag, h2CloseTag + case 3: + return h3Tag, h3CloseTag + case 4: + return h4Tag, h4CloseTag + case 5: + return h5Tag, h5CloseTag + default: + return h6Tag, h6CloseTag + } +} + +func (r *HTMLRenderer) outHRTag(w io.Writer) { + if r.Flags&UseXHTML == 0 { + r.out(w, hrTag) + } else { + r.out(w, hrXHTMLTag) + } +} + // RenderNode is a default renderer of a single node of a syntax tree. For // block nodes it will be called twice: first time with entering=true, second // time with entering=false, so that it could know when it's working on an open@@ -420,25 +513,29 @@ case Softbreak:
r.out(w, []byte{'\n'}) // TODO: make it configurable via out(renderer.softbreak) case Hardbreak: - r.out(w, tag("br", nil, true)) + if r.Flags&UseXHTML == 0 { + r.out(w, brTag) + } else { + r.out(w, brXHTMLTag) + } r.cr(w) case Emph: if entering { - r.out(w, tag("em", nil, false)) + r.out(w, emTag) } else { - r.out(w, tag("/em", nil, false)) + r.out(w, emCloseTag) } case Strong: if entering { - r.out(w, tag("strong", nil, false)) + r.out(w, strongTag) } else { - r.out(w, tag("/strong", nil, false)) + r.out(w, strongCloseTag) } case Del: if entering { - r.out(w, tag("del", nil, false)) + r.out(w, delTag) } else { - r.out(w, tag("/del", nil, false)) + r.out(w, delCloseTag) } case HTMLSpan: if r.Flags&SkipHTML != 0 {@@ -457,9 +554,9 @@ // mark it but don't link it if it is not a safe link: no smartypants
dest := node.LinkData.Destination if needSkipLink(r.Flags, dest) { if entering { - r.out(w, tag("tt", nil, false)) + r.out(w, ttTag) } else { - r.out(w, tag("/tt", nil, false)) + r.out(w, ttCloseTag) } } else { if entering {@@ -475,12 +572,12 @@ attrs = appendLinkAttrs(attrs, r.Flags, dest)
if len(node.LinkData.Title) > 0 { attrs = append(attrs, fmt.Sprintf("title=%q", esc(node.LinkData.Title))) } - r.out(w, tag("a", attrs, false)) + r.tag(w, aTag, attrs) } else { if node.NoteID != 0 { break } - r.out(w, tag("/a", nil, false)) + r.out(w, aCloseTag) } } case Image:@@ -509,9 +606,9 @@ r.out(w, []byte(`" />`))
} } case Code: - r.out(w, tag("code", nil, false)) + r.out(w, codeTag) r.out(w, escCode(node.Literal)) - r.out(w, tag("/code", nil, false)) + r.out(w, codeCloseTag) case Document: break case Paragraph:@@ -530,9 +627,9 @@ }
if node.Parent.Type == BlockQuote && node.Prev == nil { r.cr(w) } - r.out(w, tag("p", attrs, false)) + r.out(w, pTag) } else { - r.out(w, tag("/p", attrs, false)) + r.out(w, pCloseTag) if !(node.Parent.Type == Item && node.Next == nil) { r.cr(w) }@@ -540,9 +637,9 @@ }
case BlockQuote: if entering { r.cr(w) - r.out(w, tag("blockquote", attrs, false)) + r.out(w, blockquoteTag) } else { - r.out(w, tag("/blockquote", nil, false)) + r.out(w, blockquoteCloseTag) r.cr(w) } case HTMLBlock:@@ -553,7 +650,7 @@ r.cr(w)
r.out(w, node.Literal) r.cr(w) case Header: - tagname := fmt.Sprintf("h%d", node.Level) + openTag, closeTag := headerTagsFromLevel(node.Level) if entering { if node.IsTitleblock { attrs = append(attrs, `class="title"`)@@ -569,39 +666,42 @@ }
attrs = append(attrs, fmt.Sprintf(`id="%s"`, id)) } r.cr(w) - r.out(w, tag(tagname, attrs, false)) + r.tag(w, openTag, attrs) } else { - r.out(w, tag("/"+tagname, nil, false)) + r.out(w, closeTag) if !(node.Parent.Type == Item && node.Next == nil) { r.cr(w) } } case HorizontalRule: r.cr(w) - r.out(w, tag("hr", attrs, r.Flags&UseXHTML != 0)) + r.outHRTag(w) r.cr(w) case List: - tagName := "ul" + openTag := ulTag + closeTag := ulCloseTag if node.ListFlags&ListTypeOrdered != 0 { - tagName = "ol" + openTag = olTag + closeTag = olCloseTag } if node.ListFlags&ListTypeDefinition != 0 { - tagName = "dl" + openTag = dlTag + closeTag = dlCloseTag } if entering { if node.IsFootnotesList { - r.out(w, []byte("\n<div class=\"footnotes\">\n\n")) - r.out(w, tag("hr", attrs, r.Flags&UseXHTML != 0)) + r.out(w, footnotesDivBytes) + r.outHRTag(w) r.cr(w) } r.cr(w) if node.Parent.Type == Item && node.Parent.Parent.Tight { r.cr(w) } - r.out(w, tag(tagName, attrs, false)) + r.tag(w, openTag[:len(openTag)-1], attrs) r.cr(w) } else { - r.out(w, tag("/"+tagName, nil, false)) + r.out(w, closeTag) //cr(w) //if node.parent.Type != Item { // cr(w)@@ -613,16 +713,19 @@ if node.Parent.Type == Document || node.Parent.Type == BlockQuote {
r.cr(w) } if node.IsFootnotesList { - r.out(w, []byte("\n</div>\n")) + r.out(w, footnotesCloseDivBytes) } } case Item: - tagName := "li" + openTag := liTag + closeTag := liCloseTag if node.ListFlags&ListTypeDefinition != 0 { - tagName = "dd" + openTag = ddTag + closeTag = ddCloseTag } if node.ListFlags&ListTypeTerm != 0 { - tagName = "dt" + openTag = dtTag + closeTag = dtCloseTag } if entering { if itemOpenCR(node) {@@ -633,7 +736,7 @@ slug := slugify(node.ListData.RefLink)
r.out(w, footnoteItem(r.FootnoteAnchorPrefix, slug)) break } - r.out(w, tag(tagName, nil, false)) + r.out(w, openTag) } else { if node.ListData.RefLink != nil { slug := slugify(node.ListData.RefLink)@@ -641,32 +744,34 @@ if r.Flags&FootnoteReturnLinks != 0 {
r.out(w, footnoteReturnLink(r.FootnoteAnchorPrefix, r.FootnoteReturnLinkContents, slug)) } } - r.out(w, tag("/"+tagName, nil, false)) + r.out(w, closeTag) r.cr(w) } case CodeBlock: attrs = appendLanguageAttr(attrs, node.Info) r.cr(w) - r.out(w, tag("pre", nil, false)) - r.out(w, tag("code", attrs, false)) + r.out(w, preTag) + r.tag(w, codeTag[:len(codeTag)-1], attrs) r.out(w, escCode(node.Literal)) - r.out(w, tag("/code", nil, false)) - r.out(w, tag("/pre", nil, false)) + r.out(w, codeCloseTag) + r.out(w, preCloseTag) if node.Parent.Type != Item { r.cr(w) } case Table: if entering { r.cr(w) - r.out(w, tag("table", nil, false)) + r.out(w, tableTag) } else { - r.out(w, tag("/table", nil, false)) + r.out(w, tableCloseTag) r.cr(w) } case TableCell: - tagName := "td" + openTag := tdTag + closeTag := tdCloseTag if node.IsHeader { - tagName = "th" + openTag = thTag + closeTag = thCloseTag } if entering { align := cellAlignment(node.Align)@@ -676,37 +781,37 @@ }
if node.Prev == nil { r.cr(w) } - r.out(w, tag(tagName, attrs, false)) + r.tag(w, openTag, attrs) } else { - r.out(w, tag("/"+tagName, nil, false)) + r.out(w, closeTag) r.cr(w) } case TableHead: if entering { r.cr(w) - r.out(w, tag("thead", nil, false)) + r.out(w, theadTag) } else { - r.out(w, tag("/thead", nil, false)) + r.out(w, theadCloseTag) r.cr(w) } case TableBody: if entering { r.cr(w) - r.out(w, tag("tbody", nil, false)) + r.out(w, tbodyTag) // XXX: this is to adhere to a rather silly test. Should fix test. if node.FirstChild == nil { r.cr(w) } } else { - r.out(w, tag("/tbody", nil, false)) + r.out(w, tbodyCloseTag) r.cr(w) } case TableRow: if entering { r.cr(w) - r.out(w, tag("tr", nil, false)) + r.out(w, trTag) } else { - r.out(w, tag("/tr", nil, false)) + r.out(w, trCloseTag) r.cr(w) } default: