Go

Build htomd from source and extract Markdown and metadata with Go.

Requires Go 1.27 or later. No external dependencies.

Install from source

The Go port is available in the repository; it does not yet have a module release. Clone the repository and use a local replacement in your Go project:

git clone https://github.com/jamiedavenport/htomd.git
# Run these in your application's Go module, adjusting the checkout path:
go mod edit -replace=github.com/jamiedavenport/htomd/go=./htomd/go
go get github.com/jamiedavenport/htomd/go

To install the CLI from the checkout:

go install -C htomd/go ./cmd/htomd

Quickstart

package main

import (
    "fmt"
    "log"

    "github.com/jamiedavenport/htomd/go"
)

func main() {
    html := "<article><h1>Tea</h1><p>Steep gently.</p></article>"
    source := "https://example.org/tea"
    document, err := htomd.Extract(html, htomd.Options{URL: &source})
    if err != nil {
        log.Fatal(err)
    }
    fmt.Print(document.Markdown) // "# Tea\n\nSteep gently.\n"
    fmt.Println(*document.Metadata.Title) // "Tea"
}

API

func Convert(html string, options Options) (string, error)
func Extract(html string, options Options) (Document, error)

Use Options{} without a source URL. Options.URL is a *string; nil means absent and a pointer to "" preserves an empty source URL. HTML and the supplied URL must be valid UTF-8; otherwise the functions return an error. Malformed HTML is recovered best-effort. No URL is fetched.

Results own their data and may be modified. Check metadata pointers for nil before dereferencing them. Strategy constants are Semantic, Scored, Fallback, and None. JSON serialization uses snake_case field names and null for absent metadata.

Native behavior

Go uses unicode.IsSpace, html.UnescapeString, net/url, and encoding/json. List counters accept ASCII decimal digits with an optional sign. See the package README for the resulting differences from Python.

See shared behavior and limitations and the CLI.