.

brson · May 8, 2020 · 581211d · 581211d
1 parent f9b4681
commit 581211d
Show file tree

Hide file tree

Showing 4 changed files with 49 additions and 8 deletions.
diff --git a/src/convert.rs b/src/convert.rs
@@ -3,7 +3,7 @@ use std::mem;
 use anyhow::Result;
 use crate::html::SubDom;
 use markup5ever_rcdom as rcdom;
-use rcdom::{Node as Node, NodeData};
+use rcdom::{Node, NodeData};
 use crate::doc;
 use log::{warn, debug, error};
 use crate::config::BlogPost;

diff --git a/src/doc.rs b/src/doc.rs
@@ -32,7 +32,7 @@ pub struct Heading {
     pub level: HeadingLevel,
 }
 
-#[derive(Debug, Clone, Copy)]
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
 pub enum HeadingLevel {
     H1, H2, H3, H4, H5, H6,
 }

diff --git a/src/main.rs b/src/main.rs
@@ -263,7 +263,7 @@ fn run_render_article(cmd: CmdOpts<RenderArticle>) -> Result<()> {
         match html::extract_article(&post) {
             Ok(dom) => {
                 let doc = convert::from_dom(&meta, &dom);
-                let doc = sanitize::sanitize(doc);
+                let doc = sanitize::sanitize(doc, &dom);
                 let title = extract::title(&doc);
                 let doc = render::to_string(&assets, &doc)?;
                 if !cmd.cmd.to_file {
@@ -310,7 +310,7 @@ fn run_extract_title(cmd: CmdOpts<ExtractTitle>) -> Result<()> {
         match html::extract_article(&post) {
             Ok(dom) => {
                 let doc = convert::from_dom(&meta, &dom);
-                let doc = sanitize::sanitize(doc);
+                let doc = sanitize::sanitize(doc, &dom);
                 let title = extract::title(&doc);
                 match title {
                     Some(title) => {
@@ -334,7 +334,7 @@ fn run_generate_slug(cmd: CmdOpts<GenerateSlug>) -> Result<()> {
         match html::extract_article(&post) {
             Ok(dom) => {
                 let doc = convert::from_dom(&meta, &dom);
-                let doc = sanitize::sanitize(doc);
+                let doc = sanitize::sanitize(doc, &dom);
                 let title = extract::title(&doc);
                 match title {
                     Some(title) => {
@@ -364,7 +364,7 @@ fn run_write_index(cmd: CmdOpts<WriteIndex>) -> Result<()> {
         match html::extract_article(&post) {
             Ok(dom) => {
                 let doc = convert::from_dom(&meta, &dom);
-                let doc = sanitize::sanitize(doc);
+                let doc = sanitize::sanitize(doc, &dom);
                 let title = extract::title(&doc);
                 match title {
                     Some(title) => {
@@ -410,7 +410,7 @@ fn run_write_author_pages(cmd: CmdOpts<WriteAuthorPages>) -> Result<()> {
         match html::extract_article(&post) {
             Ok(dom) => {
                 let doc = convert::from_dom(&meta, &dom);
-                let doc = sanitize::sanitize(doc);
+                let doc = sanitize::sanitize(doc, &dom);
                 let title = extract::title(&doc);
                 match title {
                     Some(title) => {

diff --git a/src/sanitize.rs b/src/sanitize.rs
@@ -1,7 +1,13 @@
+use log::warn;
 use anyhow::Result;
 use crate::doc::*;
+use markup5ever_rcdom as rcdom;
+use rcdom::{Node, NodeData};
+use crate::html::SubDom;
+use crate::doc::{Block, HeadingLevel};
 
-pub fn sanitize(doc: Document) -> Document {
+pub fn sanitize(doc: Document, dom: &SubDom) -> Document {
+    let doc = maybe_add_h1(doc, dom);
     doc
 }
 
@@ -65,3 +71,38 @@ fn remove_leading_and_trailing_dashes(mut s: String) -> String {
     }
     s.to_string()
 }
+
+/// Some blogs don't put their h1 title inside the `article` tag (e.g.
+/// burntsushi). This hack looks for cases where there's the extracted doc
+/// contains no h1 before other headers, then looks for an h1 inside the dom and
+/// stuff it into the doc.
+fn maybe_add_h1(mut doc: Document, dom: &SubDom) -> Document {
+    if missing_h1(&doc) {
+        warn!("missing h1 in {:?}", doc.meta.origin_url);
+        if let Some(h1) = find_h1(dom) {
+            doc.body.blocks.insert(0, Block::Heading(h1));
+        }
+    }
+    doc
+}
+
+fn missing_h1(doc: &Document) -> bool {
+    for body in &doc.body.blocks {
+        match body {
+            Block::Heading(h) => {
+                if h.level != HeadingLevel::H1 {
+                    return true;
+                } else {
+                    return false;
+                }
+            }
+            _ => { }
+        }
+    }
+
+    false
+}
+
+fn find_h1(dom: &SubDom) -> Option<Heading> {
+    None
+}