Conversation
Previous positions 15/38/62/85 had: - intra-cell row spacing: 23-24% per gap - inter-cell row spacing: (100-85) + 15 = 30% per gap That 6-7% disproportion produced a visible horizontal stripe between adjacent rows of Braille — the column of Braille cells read as discrete rows instead of a continuous grid. New positions 12.5/37.5/62.5/87.5 give uniform 25% spacing both within a cell and across cells, so a column of full-dot Braille cells flows as a single uniform grid with no visible row break. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Previous radius (0.07em) gave a 14% dot diameter relative to the 1em cell height, leaving an 11% visible gap between adjacent dots vertically (with the new 25% uniform center-to-center spacing from the prior commit). Bumped radius to 0.125em (= 25% diameter, exactly the center-to- center spacing) so adjacent dots just touch vertically. The 0.01em soft-transition zone is preserved (transparent transition at 0.135em). Visual effect: a column of full-dot Braille cells now reads as a solid vertical strip of touching dots, with no visible row-break stripe. Horizontal spacing between left and right column dots is unchanged (still 50% center-to-center), so dots in a row remain visually distinct as paired left/right columns. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Revert the radius bump from a485679 (user wants normal-sized dots, not bigger ones) and instead shift dot positions further apart vertically — pushing the top and bottom rows close to cell edges. New positions: y = 5%, 35%, 65%, 95% - intra-cell row spacing: 30% per gap - inter-cell row spacing: (100-95) + 5 = 10% per gap Intra-cell is now wider but each cell still has dots. Inter-cell is much narrower so the "empty stripe" between rows of cells visually shrinks. Spacing is no longer perfectly uniform, but the visible row- break gap is reduced from 25% → 10%, which is the actual perceived issue ("the inter-cell area looks emptier than the intra-cell area"). Dot radius stays at 0.07em (no bigger dots). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Previous edge-near positions (5/35/65/95) reduced the inter-cell gap to ~1.66px but still left a visible thin stripe between rows because the painted dots didn't quite reach the cell boundary. Pinning the top and bottom rows at y=0% and y=100% places the dot centres exactly on the cell edges. The radial-gradient only paints within the inline-block's bounds, so a y=0% dot shows just its bottom half, and a y=100% dot shows just its top half. When two cells stack vertically, the top-half from cell N+1's top dot and the bottom-half from cell N's bottom dot share the cell boundary line and together form one complete dot straddling the join — no visible gap between rows by construction. Intra-cell spacing grows from 30% to 33.33% per gap; visually the in-cell dot grid is now perfectly evenly spaced. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…te gap" This reverts commit 93cc8b6.
Cells previously had uniform 50% horizontal dot spacing both within and between cells, making adjacent Braille glyphs blur together horizontally. Moving the dot columns inward to x=27%/73% makes: - intra-cell horizontal spacing: 46% - inter-cell horizontal spacing: 54% The 8% asymmetry produces a faintly perceptible cell-boundary so adjacent Braille characters can be distinguished without obvious gaps that would break the contiguous-grid feel. Cell width stays at 1ch so Braille remains aligned with surrounding text. Only the dot positions within each cell change. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
_parse_header_color() accepted both 3-hex and 6-hex values for #!bg=/#!fg= headers, inconsistent with the 3-hex-only inline `Fxxx`/`Bxxx` colour tokens. Also removes README and example references to the `FTxxxxxx`/`BTxxxxxx` 24-bit inline extension, which was actually dropped from the parser back in v1.0.2.
# Conflicts: # CHANGELOG.md
Dev's tip was going unverified between pushes and promotion to main since ci.yml only triggered on main push/PR. security.yml stays main+PR-only per the dev-branch workflow's guidance that heavier scans can wait for the promotion gate.
Fetched and read NomadNet's own Guide.py (in-app spec) and MicronParser.py (reference implementation) directly from the markqvist/NomadNet repo to re-audit every "MeshChat parity" claim in the codebase against the real target: rendering Micron the way actual NomadNet renders it, not just matching one third-party client. Behavior fixes (both narrow, matching NomadNet exactly): - Dividers only get a custom repeat-character/double-rule treatment when the line is exactly "-" + one more char, matching MicronParser.py's `len(line) == 2` check. `---`/`-==` now correctly fall back to the default rule instead of being misread as styled. - Page-level #!bg=/#!fg= headers accept both 3-hex and 6-hex again (Guide.py doesn't restrict header length the way it restricts the inline `Fxxx`/`Bxxx` tags) — reverts an over-correction from earlier this session that assumed headers should match inline's 3-hex-only rule. Docs corrections (README, CHANGELOG, code comments): - Renamed "Dynamic include" to "Partials" (NomadNet's own term) and documented what's actually discarded (refresh/fields/pid) instead of implying they're used. - Documented pipe-separated link field-specs as the real primary syntax; the existing backtick-joining leniency is MeshChat/ Micron2HTML-only and kept but no longer presented as canonical. - Documented NomadNet's own canonical checkbox/radio style (empty field label + trailing text) alongside the already-documented embedded-label form. - Fixed broken checkbox/radio examples in README and examples/showcase.mu (missing the mandatory backtick separator). - Fixed inline colour doubling example (was missing the command letter: "F40" -> "`FF40"). - Fixed incorrect file:// wording in the Security section (the code blocks a `/file/` path segment, not the file:// URI scheme). - Fixed stale "pure Python 3.9+" claim (>=3.10 since v1.0.8). - Added README "Known limitations" section: tables and anchors are unimplemented; alignment-anywhere and inline literal-toggle are documented as intentional supersets of NomadNet's line-scoped versions. - Documented to_text()/--format text, previously untested-in-docs. Test coverage: - 2 new divider tests locking in the length-2 rule. - 1 new header-color test confirming 6-hex support. - Renamed TestDynamicInclude -> TestPartials; added a test confirming fields/pid are discarded, matching corrected docs. - Relabeled "MeshChat parity" comments to "NomadNet" throughout converter.py and tests where the behavior is independently verified against MicronParser.py, without changing behavior. 57 tests passing (was 54).
Reverts the header 6-hex acceptance added earlier this session. NomadNet's own docs would technically permit 6-hex there, but with no marker distinguishing 3-hex from 6-hex (unlike the inline tags, which would need something like NomadNet's `T` prefix), allowing both means a value's meaning silently depends on its length. One fixed width, applied consistently to every colour value in this converter, beats matching every length NomadNet's own code happens to tolerate.
…ences
Full NomadNet parity pass: implements the two previously-unimplemented
real NomadNet features (tables, anchors) and reverses several
Micron2HTML-only leniencies/extensions that had no equivalent in real
NomadNet, so a page renders the same way in both directions.
New features:
- Tables (`t): rendered as box-drawing-character ASCII art (matching
NomadNet's actual visual output, not a semantic <table>), ported from
RNS's MarkdownToMicron.format_table_raw algorithm. Supports header/
alignment-separator/data rows, per-column alignment, min column width,
optional `tc30-style whole-table alignment + max-width, escaped pipes.
- Anchors: heading auto-anchors (slugified text -> id=), explicit `:name
declarations (zero-width, first-wins shared namespace), `[label`#name]
named jumps, and bare `[label`#] "jump to next heading" links (resolved
via a one-time forward/backward pre-pass in convert()).
Behavior changes (reversing MeshChat-only leniencies, matching NomadNet
exactly):
- Removed the inline mid-line `=...`= literal toggle - NomadNet only
recognizes `= as a whole-line block toggle. This was an undocumented
Micron2HTML extension with 2 tests exercising it; both rewritten.
- Link field-specs with more than 3 backtick-separated segments now
render nothing at all, reversing a v1.0.3 fix that was solving a
MeshChat-specific quirk rather than real NomadNet behavior. The
correct pipe-separated form (`[label`url`a=1|b=2]) is unaffected.
- Empty heading lines now emit nothing (no row), matching NomadNet's
parse_line() returning None - reverses a deliberate deviation made
earlier this session, superseded by the broader parity goal.
Partials (`{url`refresh`fields}) now expose refresh/fields/pid as
data-refresh/data-fields/data-pid attributes instead of discarding them,
so a consuming app can wire up its own live-refresh behavior - still no
JS shipped by this library.
Also corrected a factually wrong README claim: "alignment tags must
appear at line start" was never an enforced NomadNet parser rule, just
Guide.py's style advice to authors - Micron2HTML already matched real
NomadNet's permissive behavior here.
Test suite: 57 -> 90 (33 new/rewritten tests). README gains Tables and
Anchors sections and an updated Known Limitations list (wcwidth and
table-shrink-formula approximations, partials live-refresh gap, a rare
anchor-collision edge case in the bare-hash pre-pass).
_close_innermost()/_pop_tag() found and removed a tag from the middle of the open-tag stack when it wasn't actually innermost, but still emitted the close tag at the current output position — HTML closes are LIFO, so that close landed on whatever tag genuinely was innermost instead, desyncing the bookkeeping from the real DOM for the rest of the line (e.g. `B777 X`f `F975`b <> left the background open until end-of-line instead of closing at `b). Fixes by unwinding: close every tag above the target (innermost-first), close the target, then reopen the unwound tags as fresh elements. The tag stack now carries each entry's open HTML alongside its close HTML so the reopens can be replayed. `!/`_/`* now route through _close_innermost() like `b/`f always did, instead of inlining their own close+pop; _pop_tag() is removed as dead code. Reported against a downstream consumer (vscode-mu-preview) as "background color not resetting after `b token" — confirmed the bug was entirely upstream here. 4 new tests in TestTagNesting.
NomadNet's reference MicronParser.py accepts an FT<6hex> / BT<6hex> form for exact 24-bit color. Micron2HTML dropped this in an earlier version citing MeshChat compatibility — MeshChat's MicronParser.js doesn't render it, so pages relying on the T escape look wrong there. Re-added now for feature parity with NomadNet's own parser; portability trade-off is documented in the docstring and CHANGELOG so authors know when it's safe to use vs when to stick to `Fxxx / `Bxxx. Invalid or too-short input consumes T + 6 chars so garbage doesn't leak as text, matching the 3-hex parser's "always consume, apply only when valid" convention. T-prefix branch checks BEFORE the 3-hex branch so FTrrggbb isn't mis-consumed as FT? + garbage. Test: existing test_24bit_T_format_not_supported flipped to test_24bit_T_format_fg (now asserts the color IS applied); added tests for the BG form, invalid-hex consumption, and the too-short (`FTab at EOF) edge case. All 97 tests pass. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
NomadNet's reference MicronParser.py accepts this form; Micron2HTML had dropped it in an earlier version citing MeshChat compat. Users authoring for Micron2HTML-based clients can now use exact colors; docstring + CHANGELOG note the portability trade-off explicitly so authors know when it's safe. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Same phantom-conflict pattern the NomadPortal release cycle hits (see [[rebase-merge-phantom-conflicts]] memory). main had the squashed v1.1.0 commit but dev still carried its individual pre-1.1.0 history. Merged with -X ours so dev's content wins on the phantom-conflict files; CHANGELOG needed a manual dedupe pass to drop the duplicated [1.1.0] section the merge produced.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Promotes dev to main for v1.1.1.
Summary
Small patch release restoring the
`FT<6hex>/`BT<6hex>24-bit color parser that NomadNet's reference implementation supports.What's in
Test plan