Require refuses an empty value. Heading ids: an explicit {#id} must be sound
and unique, generated ids avoid the explicit ones and each other, and a heading
with no ASCII letters gets a fallback. Fenced code blocks keep their
placeholders and a backslash keeps one anywhere. The em dash rule covers the
header of the manifesto pages and the HTML entities. The feed makes src links
absolute, base_url is required, and the news page must be a folder index. A
category written twice in a post counts once, a block name in a news title is no
longer taken for a placeholder, and build refuses extra arguments.
209 lines
6.2 KiB
Go
209 lines
6.2 KiB
Go
// SPDX-License-Identifier: GPL-3.0-or-later
|
|
|
|
package content
|
|
|
|
import (
|
|
"fmt"
|
|
"os"
|
|
"path/filepath"
|
|
"regexp"
|
|
"sort"
|
|
"strings"
|
|
"time"
|
|
|
|
"git.kmalloc.org/squid/motorino/internal/header"
|
|
"git.kmalloc.org/squid/motorino/internal/mdx"
|
|
)
|
|
|
|
// Post is a news item.
|
|
type Post struct {
|
|
// Slug is the name of the file without its extension.
|
|
Slug string
|
|
// Path is the address, /news/YYYY/MM/<slug>.html.
|
|
Path string
|
|
// Title, Author and Description come from the header; the description
|
|
// falls back to the first sentence of the body.
|
|
Title, Author, Description string
|
|
// Date is the date as written, with its own offset.
|
|
Date time.Time
|
|
// Categories are the categories of the post, in the order written.
|
|
Categories []Category
|
|
// Source is the file the post was read from, relative to the site.
|
|
Source string
|
|
// Body is the Markdown after the header.
|
|
Body string
|
|
}
|
|
|
|
// postKeys are the keys a news header may have.
|
|
var postKeys = []string{"Title", "Author", "Date", "Categories", "Description"}
|
|
|
|
// dateLayouts are the accepted spellings of Date, ISO 8601 with an offset.
|
|
var dateLayouts = []string{"2006-01-02T15:04:05-0700", "2006-01-02T15:04:05Z07:00"}
|
|
|
|
// slugName matches a file name made of lowercase words joined by hyphens.
|
|
var slugName = regexp.MustCompile(`^[a-z0-9]+(?:-[a-z0-9]+)*$`)
|
|
|
|
// PostPath returns the address of a news item: the year and the month are
|
|
// those of date as written, not converted to another time zone.
|
|
func PostPath(date time.Time, slug string) string {
|
|
return "/news/" + date.Format("2006/01") + "/" + slug + ".html"
|
|
}
|
|
|
|
// LoadPosts reads news/*.md, newest first. A site without a news folder has
|
|
// no posts. Two posts with the same address, or two category names with the
|
|
// same address, are errors.
|
|
func LoadPosts(dir string) ([]Post, error) {
|
|
files, err := filepath.Glob(filepath.Join(dir, "news", "*.md"))
|
|
if err != nil {
|
|
return nil, fmt.Errorf("list news: %w", err)
|
|
}
|
|
var posts []Post
|
|
for _, file := range files {
|
|
post, err := loadPost(file)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
posts = append(posts, post)
|
|
}
|
|
sort.SliceStable(posts, func(i, j int) bool {
|
|
if !posts[i].Date.Equal(posts[j].Date) {
|
|
return posts[i].Date.After(posts[j].Date)
|
|
}
|
|
return posts[i].Slug < posts[j].Slug
|
|
})
|
|
if err := checkPaths(posts); err != nil {
|
|
return nil, err
|
|
}
|
|
if err := checkCategories(posts); err != nil {
|
|
return nil, err
|
|
}
|
|
return posts, nil
|
|
}
|
|
|
|
// loadPost reads and checks one news file.
|
|
func loadPost(file string) (Post, error) {
|
|
base := filepath.Base(file)
|
|
source := "news/" + base
|
|
slug := strings.TrimSuffix(base, ".md")
|
|
if !slugName.MatchString(slug) {
|
|
return Post{}, fmt.Errorf("%s: the name must be lowercase words joined by hyphens", source)
|
|
}
|
|
data, err := os.ReadFile(file)
|
|
if err != nil {
|
|
return Post{}, fmt.Errorf("read %s: %w", source, err)
|
|
}
|
|
h, body, err := header.Parse(source, data)
|
|
if err != nil {
|
|
return Post{}, err
|
|
}
|
|
if err := h.Only(source, postKeys...); err != nil {
|
|
return Post{}, err
|
|
}
|
|
if err := h.Require(source, "Title", "Date"); err != nil {
|
|
return Post{}, err
|
|
}
|
|
if err := mdx.RejectEmDash(source, string(data)); err != nil {
|
|
return Post{}, err
|
|
}
|
|
date, err := parseDate(source, h.Get("Date"))
|
|
if err != nil {
|
|
return Post{}, err
|
|
}
|
|
categories, err := parseCategories(source, h.Get("Categories"))
|
|
if err != nil {
|
|
return Post{}, err
|
|
}
|
|
description := h.Get("Description")
|
|
if description == "" {
|
|
description = firstSentence(body)
|
|
}
|
|
return Post{
|
|
Slug: slug, Path: PostPath(date, slug), Title: h.Get("Title"), Author: h.Get("Author"),
|
|
Description: description, Date: date, Categories: categories, Source: source, Body: body,
|
|
}, nil
|
|
}
|
|
|
|
// parseDate reads Date in any of the accepted layouts.
|
|
func parseDate(source, value string) (time.Time, error) {
|
|
for _, layout := range dateLayouts {
|
|
if date, err := time.Parse(layout, value); err == nil {
|
|
return date, nil
|
|
}
|
|
}
|
|
return time.Time{}, fmt.Errorf("%s: Date %q is not ISO 8601 with an offset, such as 2026-09-18T17:00:45+0200", source, value)
|
|
}
|
|
|
|
// parseCategories splits a comma separated list into categories.
|
|
func parseCategories(source, value string) ([]Category, error) {
|
|
var out []Category
|
|
seen := map[string]bool{}
|
|
for _, name := range strings.Split(value, ",") {
|
|
name = strings.TrimSpace(name)
|
|
if name == "" {
|
|
continue
|
|
}
|
|
slug, err := CategorySlug(name)
|
|
if err != nil {
|
|
return nil, fmt.Errorf("%s: %w", source, err)
|
|
}
|
|
if seen[slug] {
|
|
continue
|
|
}
|
|
seen[slug] = true
|
|
out = append(out, Category{Name: name, Slug: slug})
|
|
}
|
|
return out, nil
|
|
}
|
|
|
|
// checkPaths returns an error when two posts have the same address.
|
|
func checkPaths(posts []Post) error {
|
|
seen := map[string]string{}
|
|
for _, p := range posts {
|
|
if other, dup := seen[p.Path]; dup {
|
|
return fmt.Errorf("%s and %s have the same address %s", other, p.Source, p.Path)
|
|
}
|
|
seen[p.Path] = p.Source
|
|
}
|
|
return nil
|
|
}
|
|
|
|
// checkCategories returns an error when two different category names give the
|
|
// same address.
|
|
func checkCategories(posts []Post) error {
|
|
names := map[string]string{}
|
|
for _, p := range posts {
|
|
for _, c := range p.Categories {
|
|
if other, ok := names[c.Slug]; ok && other != c.Name {
|
|
return fmt.Errorf("%s: the categories %q and %q give the same address /news/category/%s.html", p.Source, other, c.Name, c.Slug)
|
|
}
|
|
names[c.Slug] = c.Name
|
|
}
|
|
}
|
|
return nil
|
|
}
|
|
|
|
// markdownLink, markdownMark and sentenceEnd pick the first sentence out of a
|
|
// body and drop its Markdown markup, for a description that was not written.
|
|
var (
|
|
markdownLink = regexp.MustCompile(`\[([^\]]*)\]\([^)]*\)`)
|
|
markdownMark = regexp.MustCompile("[*_`]")
|
|
sentenceEnd = regexp.MustCompile(`^(.+?[.!?])(?:\s|$)`)
|
|
)
|
|
|
|
// firstSentence returns the first sentence of the first paragraph of body that
|
|
// is not a heading or raw HTML, without Markdown markup.
|
|
func firstSentence(body string) string {
|
|
for _, block := range strings.Split(body, "\n\n") {
|
|
block = strings.TrimSpace(block)
|
|
if block == "" || strings.HasPrefix(block, "#") || strings.HasPrefix(block, "<") {
|
|
continue
|
|
}
|
|
text := markdownLink.ReplaceAllString(strings.Join(strings.Fields(block), " "), "$1")
|
|
text = markdownMark.ReplaceAllString(text, "")
|
|
if m := sentenceEnd.FindStringSubmatch(text); m != nil {
|
|
return m[1]
|
|
}
|
|
return text
|
|
}
|
|
return ""
|
|
}
|