Documentation
¶
Overview ¶
Code generated from @primer/octicons 19.15.1 (MIT). DO NOT EDIT by hand.
Package viewer renders blob content for the file viewer page. Each file type is one Renderer behind a small interface; the server asks the Registry and never learns what file types exist. Structural blob states (symlink, LFS pointer, too big, empty) stay in the server: they are states of the record, not content (spec 2011 doc 18).
Index ¶
- func ChromaCSS() string
- func FindReadme(entries []backend.TreeEntry) (string, bool)
- func HighlightLines(filename, content string) []template.HTML
- func IsMarkdownFile(name string) bool
- func MIME(ext string) (string, bool)
- func Octicon(name string) template.HTML
- func RenderMarkdown(mctx MarkdownContext, src []byte) template.HTML
- type Input
- type MarkdownContext
- type Output
- type Registry
- type Renderer
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func ChromaCSS ¶
func ChromaCSS() string
ChromaCSS renders the github light style plus the github-dark style scoped to the dark color mode, with an auto-mode media query copy.
func FindReadme ¶
FindReadme picks the readme github would show for a directory listing.
func HighlightLines ¶
HighlightLines tokenizes content and returns one HTML fragment per line, using chroma class names so the github and github-dark styles switch with the page theme instead of re-rendering.
func IsMarkdownFile ¶
IsMarkdownFile reports whether the blob should get a rendered preview.
func Octicon ¶
Octicon returns the inline SVG for a named icon, "" markup for unknown names. Page templates use this through the server's octicon func; the markdown pipeline uses the map directly.
func RenderMarkdown ¶
func RenderMarkdown(mctx MarkdownContext, src []byte) template.HTML
RenderMarkdown runs the pipeline: goldmark with GFM and our transformers, then bluemonday, then trusted post-passes that add heading anchors and alert icons. Sanitizing after rendering is the security boundary; the post-passes only touch markup the sanitizer already approved.
Types ¶
type Input ¶
type Input struct {
Name string // base name, "model.parquet"
Path string // repo-relative path, "data/model.parquet"
Ext string // lowercased extension with dot, ".parquet"
Content []byte
Size int64 // true blob size, equals len(Content) here
Binary bool // backend's NUL-byte verdict
RawURL string // "/{key}/raw/{ref}/{path}", for media and frame sources
RepoPath string // "/{owner}/{repo}", for markdown link rewriting
Ref string // resolved display ref, same purpose
Plain bool // ?plain=1: the user asked for source, not preview
}
Input is everything a renderer may look at. Content is the full blob, already capped at backend.MaxBlobBytes by the backend; TooBig blobs never reach the registry.
type MarkdownContext ¶
type MarkdownContext struct {
RepoPath string // "/owner/name"
Ref string
Dir string // directory of the source file, "" at the root
}
MarkdownContext tells the rewriter where the markdown file lives so relative links and images resolve to blob and raw URLs.
type Output ¶
type Output struct {
Kind string // stable id: "code", "csv", "video", ...
Body template.HTML // the content area, safe by construction
Info string // toolbar text: "183 lines", "100 rows x 12 columns"
Toggle bool // offer the Preview / Code segmented control
Note string // cap warning shown above the body, "" for none
}
Output is a rendered fragment plus toolbar facts.
type Registry ¶
type Registry struct {
// contains filtered or unexported fields
}
Registry is an ordered list of renderers; the first Match wins.
func NewRegistry ¶
func (*Registry) Probe ¶
Probe reports the kind that would claim the input, without rendering. The server probes with Plain unset to keep the Preview / Code toggle visible while the plain view is showing.
func (*Registry) Render ¶
Render walks renderers in order. A renderer error is a degradation, never a failure: the walk continues, so a corrupt container falls through to whatever would have claimed the file otherwise. ok reports whether anything rendered; the server shows the structural binary or empty notice when nothing did.
type Renderer ¶
Renderer is one viewer. Match must be cheap (extension checks, magic bytes); Render may parse. A Renderer that matched but cannot render returns an error and the registry falls through to the next match.
func Audio ¶
func Audio() Renderer
Audio renders an <audio> element over the raw URL. preload="metadata" keeps a directory crawl from pulling whole files; never autoplay. Codec support is the browser's problem: the element shows its native error UI when it cannot play the file.
func Code ¶
func Code() Renderer
Code is the final text fallback: the chroma table with line ids that back the L{n} anchor contract (doc 08).
func Docx ¶
func Docx() Renderer
Docx extracts readable text from a docx file: headings, paragraphs, bold and italic runs, flat lists, one level of tables, external hyperlinks. Layout fidelity is a non-goal; the bar is strictly better than a binary notice, which is what github.com shows. Every text node passes through template escaping and the only markup in the output is the fixed mapping below, so nothing unsanitized can reach the page (spec 2011 doc 18 section 12).
func Image ¶
func Image() Renderer
Image renders raster and vector images through an <img> over the raw URL. SVG stays behind <img>, where scripts never execute; the raw response behind it keeps its sandbox CSP for direct navigation.
func Markdown ¶
func Markdown() Renderer
Markdown is the doc 09 pipeline behind the registry: rendered preview by default, deferring to the code viewer under ?plain=1.
func PDF ¶
func PDF() Renderer
PDF embeds the browser's own viewer in an iframe over the raw URL. No magic check: a text file named x.pdf shows the viewer's broken document message, which is more honest than code-viewing it. The iframe carries no sandbox attribute on purpose: a sandboxed frame blocks Chromium's PDF plugin outright, and the raw response behind it already drops the CSP sandbox for application/pdf with the frame-ancestors 'self' analysis from spec 2011 doc 18 section 4.3.
func Parquet ¶
func Parquet() Renderer
Parquet renders the schema panel and the first rows of a parquet file. The schema comes first because for large datasets the schema is the document; the row count in the Note is free footer metadata, never a full scan. A corrupt or truncated file falls through to the structural binary notice per the registry's error semantics.
Source Files