chunkmd, mdscrub: the chapter a converted book is actually divided by

Two things between a converted EPUB and its chapter titles.

The heading level taken as the chapter map was the shallowest present. Pandoc
writes the book's title as the only `#` and every real chapter as a `##`, so
the shallowest level had one heading in it and the whole book came back as a
single chapter. The dividing level is now the shallowest one used more than
once — a level used once is a title.

And an empty bracketed span survived the heading cleaner, because the pattern
wanted at least one character between the brackets. Pandoc writes an anchor
as exactly that: `## []{#link2H_INTR} Introduction`.

Gibbon's first volume, converted from the Project Gutenberg EPUB: 2 distinct
chapter titles before, 61 after, and none of them carrying an element id.

crates/chunkmd/src/pages.rs

@@ -4,6 +4,7 @@
 //! and chapter boundaries as headings. Both are needed on every chunk: a
 //! passage a reader cannot locate in the book is a passage they cannot check.
 
+use std::collections::BTreeMap;
 use std::sync::LazyLock;
 
 use regex::Regex;
@@ -117,12 +118,22 @@ pub fn chapters(markdown: &str) -> Vec<Chapter> {
         return Vec::new();
     }
 
-    let shallowest = headings
+    // The shallowest level that divides the document, which is not always the
+    // shallowest level in it. A converted EPUB carries the book's title as a
+    // lone `#` above sixty-two `##` chapters, and taking the lone one makes
+    // the whole book a single chapter — so a level used once is a title and a
+    // level used twice is a chapter map.
+    let mut counts: BTreeMap<usize, usize> = BTreeMap::new();
+    for (level, _, _) in &headings {
+        *counts.entry(*level).or_default() += 1;
+    }
+    let dividing = counts
         .iter()
-        .map(|(level, _, _)| *level)
-        .min()
+        .find(|(_, used)| **used > 1)
+        .map(|(level, _)| *level)
+        .or_else(|| counts.keys().next().copied())
         .expect("a heading");
-    headings.retain(|(level, _, _)| *level == shallowest);
+    headings.retain(|(level, _, _)| *level == dividing);
 
     let page_at = PageIndex::new(markdown);
     let mut chapters = Vec::with_capacity(headings.len());
@@ -245,6 +256,22 @@ mod tests {
         assert_eq!(pages[0].number, Some(7));
     }
 
+    #[test]
+    fn a_lone_shallow_heading_is_a_title_rather_than_a_chapter_map() {
+        // What a converted EPUB looks like: the book's title as the only
+        // `#`, and every actual chapter a `##` under it. Taking the lone
+        // one makes the whole book a single chapter.
+        let markdown = "# The Whole Book\n\n## Chapter One\n\ntext\n\n## Chapter Two\n\ntext";
+        let found = chapters(markdown);
+        assert_eq!(
+            found
+                .iter()
+                .map(|one| one.title.as_str())
+                .collect::<Vec<_>>(),
+            ["Chapter One", "Chapter Two"]
+        );
+    }
+
     #[test]
     fn chapters_come_from_the_shallowest_heading_level_present() {
         let markdown = "## Alpha\n\ntext\n\n### Sub\n\ntext\n\n## Beta\n\ntext";

crates/mdscrub/src/heading.rs

@@ -18,8 +18,10 @@ fn pattern(source: &str) -> Regex {
 /// `[text](url)` — a heading wrapped in a link to its own anchor.
 static LINK: LazyLock<Regex> = LazyLock::new(|| pattern(r"\[([^\]]*)\]\([^)]*\)"));
 
-/// `[text]{.cls}` — pandoc's bracketed inline span.
-static BRACKETED: LazyLock<Regex> = LazyLock::new(|| pattern(r"\[([^\]]+)\]\{[^}]*\}"));
+/// `[text]{.cls}` — pandoc's bracketed inline span. The text may be empty:
+/// `[]{#chapter-3}` is how an anchor with nothing in it arrives, and a
+/// heading that keeps one is a chapter title with an element id in it.
+static BRACKETED: LazyLock<Regex> = LazyLock::new(|| pattern(r"\[([^\]]*)\]\{[^}]*\}"));
 
 /// A trailing `{.attr}` block.
 static ATTRIBUTE_TAIL: LazyLock<Regex> =