Fix index sort order for IDs with differing digit counts

[linyows]

Problem

draft <template> index could place 0010 before 0009. Two defects caused this.

1. The sort field was compared byte by byte. sortDocuments used mem.order(u8, ...), so IDs written with different digit counts were ordered as text: "10" sorts before "9", and "1000" before "999".

2. The ID taken from a filename was truncated to three characters. extractDocumentMeta read filename[0..3], but the default filename_format for the adr template is {{@id{4}}}-{{@title{kebab}}}.md. A document without an - ID: line therefore got "000" from 0009-foo.md and "010" from 0100-foo.md, which is both a wrong value in the ID column and a wrong sort key.

Observed before the fix, for filenames without an - ID: line:

0009-a.md -> id='000'
0010-b.md -> id='001'
0100-c.md -> id='010'

and for IDs 9 and 10 taken from the document body, 0010-b.md sorted ahead of 0009-a.md.

Changes

• Added compareNatural, which compares runs of digits by numeric value and the surrounding text byte by byte. Values that are numerically equal but written differently ("9" and "009") fall back to a byte comparison so the ordering stays total. It is used for every string sort field, so @date and @title benefit as well.
• extractDocumentMeta now takes all leading digits of the filename as the ID, whatever the width.
• Documents whose sort field is equal are ordered by filename. Previously ties kept the directory iteration order, which is not stable across machines, so the generated index could differ between runs.

Result

With IDs 8, 9, 10, 11, 100 in the document bodies and one document that has no - ID: line (0012-no-id-line.md):

| ID | Title | Status |
|------|------|------|
| [8](./0008-doc-8.md) | Doc 8 | Proposed |
| [9](./0009-doc-9.md) | Doc 9 | Proposed |
| [10](./0010-doc-10.md) | Doc 10 | Proposed |
| [11](./0011-doc-11.md) | Doc 11 | Proposed |
| [0012](./0012-no-id-line.md) | No Id Line |  |
| [100](./0100-doc-100.md) | Doc 100 | Proposed |

Testing

• Four tests added: ID extraction from a 4-digit filename, compareNatural across digit widths and zero padding, numeric ordering in sortDocuments, and the filename tiebreaker.
• zig build test --summary all: 64/64 tests passed
• zig fmt --check reports no differences
• Manual run of the built binary on the fixture above

README documents the natural ordering and the filename tiebreaker in the index sorting section.

🤖 Generated with Claude Code

https://claude.ai/code/session_017bJFwP1iT8PKTQMWuMAYkZ