| 1 | // SPDX-License-Identifier: AGPL-3.0-or-later |
| 2 | |
| 3 | package markdown |
| 4 | |
| 5 | import ( |
| 6 | "bytes" |
| 7 | "context" |
| 8 | "errors" |
| 9 | |
| 10 | "github.com/yuin/goldmark" |
| 11 | "github.com/yuin/goldmark/extension" |
| 12 | "github.com/yuin/goldmark/parser" |
| 13 | "github.com/yuin/goldmark/renderer" |
| 14 | gmhtml "github.com/yuin/goldmark/renderer/html" |
| 15 | |
| 16 | "github.com/tenseleyFlow/shithub/internal/markdown/extensions" |
| 17 | ) |
| 18 | |
| 19 | // ErrInputTooLarge is returned when source exceeds MaxRenderInputBytes. |
| 20 | // Callers should reject the input at the API layer before reaching |
| 21 | // Render; this is the renderer's defensive fallback. |
| 22 | var ErrInputTooLarge = errors.New("markdown: source exceeds MaxRenderInputBytes") |
| 23 | |
| 24 | // Render is the canonical markdown entry point. The output bytes |
| 25 | // have already been Goldmark-rendered AND bluemonday-sanitized; |
| 26 | // callers can wrap them as `template.HTML` and inject into a |
| 27 | // template directly. |
| 28 | // |
| 29 | // `refs` and `mentions` carry the resolved cross-reference/mention |
| 30 | // state for downstream consumers (S29 notification fan-out, S21 |
| 31 | // issue_references index). The order matches first-occurrence in |
| 32 | // source. |
| 33 | // |
| 34 | // A nil ctx is fine; resolvers receive context.Background() in |
| 35 | // that case. |
| 36 | func Render(ctx context.Context, src []byte, opts Options) (rendered []byte, refs []Ref, mentions []Mention, err error) { |
| 37 | if len(src) == 0 { |
| 38 | return nil, nil, nil, nil |
| 39 | } |
| 40 | if len(src) > MaxRenderInputBytes { |
| 41 | return nil, nil, nil, ErrInputTooLarge |
| 42 | } |
| 43 | if ctx == nil { |
| 44 | ctx = context.Background() |
| 45 | } |
| 46 | |
| 47 | // Build a fresh Goldmark for each render so the per-call |
| 48 | // extension Options can plug in without races. The build is |
| 49 | // cheap (~µs) and removes any cross-render contamination of |
| 50 | // the AST transformer state. |
| 51 | xRefs := []extensions.Ref{} |
| 52 | xMentions := []extensions.Mention{} |
| 53 | xOpts := &extensions.Options{ |
| 54 | Ctx: ctx, |
| 55 | ViewerUserID: opts.ViewerUserID, |
| 56 | Refs: &xRefs, |
| 57 | Mentions: &xMentions, |
| 58 | Resolvers: extensions.Resolvers{ |
| 59 | User: opts.Resolvers.User, |
| 60 | Issue: opts.Resolvers.Issue, |
| 61 | Commit: opts.Resolvers.Commit, |
| 62 | Team: opts.Resolvers.Team, |
| 63 | }, |
| 64 | } |
| 65 | if opts.Repo != nil { |
| 66 | xOpts.RepoOwner = opts.Repo.OwnerUsername |
| 67 | xOpts.RepoName = opts.Repo.RepoName |
| 68 | } |
| 69 | |
| 70 | // gmhtml.WithUnsafe lets raw HTML through the parser → bluemonday |
| 71 | // scrubs at the sanitizer pass. Without this, every <details>, |
| 72 | // <kbd>, <sup>, etc. that users type would be HTML-escaped before |
| 73 | // the sanitizer ever sees them. The strict policy in sanitize.go |
| 74 | // is the security boundary. |
| 75 | htmlOpts := []renderer.Option{gmhtml.WithXHTML(), gmhtml.WithUnsafe()} |
| 76 | if opts.SoftBreakAsBR { |
| 77 | htmlOpts = append(htmlOpts, gmhtml.WithHardWraps()) |
| 78 | } |
| 79 | |
| 80 | gm := goldmark.New( |
| 81 | goldmark.WithExtensions( |
| 82 | extension.GFM, // tables, strikethrough, autolinks, task list |
| 83 | extensions.New(xOpts), |
| 84 | ), |
| 85 | goldmark.WithParserOptions(parser.WithAutoHeadingID()), |
| 86 | goldmark.WithRendererOptions(htmlOpts...), |
| 87 | ) |
| 88 | |
| 89 | var buf bytes.Buffer |
| 90 | if err := gm.Convert(src, &buf); err != nil { |
| 91 | return nil, nil, nil, err |
| 92 | } |
| 93 | clean := sanitizeBytes(buf.Bytes()) |
| 94 | |
| 95 | // Convert extension-local types into the public types. |
| 96 | if len(xRefs) > 0 { |
| 97 | refs = make([]Ref, len(xRefs)) |
| 98 | for i, r := range xRefs { |
| 99 | refs[i] = Ref{ |
| 100 | Kind: r.Kind, |
| 101 | Owner: r.Owner, |
| 102 | Repo: r.Repo, |
| 103 | Number: r.Number, |
| 104 | FullSHA: r.FullSHA, |
| 105 | Href: r.Href, |
| 106 | } |
| 107 | } |
| 108 | } |
| 109 | if len(xMentions) > 0 { |
| 110 | mentions = make([]Mention, len(xMentions)) |
| 111 | for i, m := range xMentions { |
| 112 | mentions[i] = Mention{Username: m.Username, Href: m.Href} |
| 113 | } |
| 114 | } |
| 115 | return clean, refs, mentions, nil |
| 116 | } |
| 117 | |
| 118 | // RenderHTML is the back-compat shim for callers that don't need |
| 119 | // resolved refs/mentions. SoftBreakAsBR defaults to true (matches |
| 120 | // the comment-style legacy behavior of the S17 helper this replaces). |
| 121 | // |
| 122 | // The signature matches `internal/repos/markdown.RenderHTML` so |
| 123 | // existing callers can swap the import path without code changes. |
| 124 | // New callers should prefer Render directly. |
| 125 | func RenderHTML(src []byte) (string, error) { |
| 126 | out, _, _, err := Render(context.Background(), src, Options{SoftBreakAsBR: true}) |
| 127 | if err != nil { |
| 128 | return "", err |
| 129 | } |
| 130 | return string(out), nil |
| 131 | } |
| 132 | |
| 133 | // RenderDocumentHTML renders README-like markdown where source line |
| 134 | // wrapping should remain semantic markdown, not comment-style hard |
| 135 | // breaks. |
| 136 | func RenderDocumentHTML(src []byte) (string, error) { |
| 137 | out, _, _, err := Render(context.Background(), src, Options{SoftBreakAsBR: false}) |
| 138 | if err != nil { |
| 139 | return "", err |
| 140 | } |
| 141 | return string(out), nil |
| 142 | } |
| 143 |