| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The task this branch of work set out to test was whether the template
engine is genuinely flavour-agnostic outside lib/render/escape.ml. It
is not, quite: bin/main.ml's own --flavour error/help text and both
man pages hand-typed the flavour list and its count ("the six
flavours") as literal prose, so adding Typst as a seventh left three
places silently wrong (an incomplete flavour list in two error
messages, and "six" where the true count is now seven) until this
commit.
bin/main.ml's two error messages (unknown --flavour value, and cannot
infer a flavour from an extension) now build their flavour list by
walking Escape.all/to_string instead of a literal string, so the next
flavour added will not reintroduce the same drift; the static --help
body text and colitur(1)/colitur-templates(5) still name flavours and
their count by hand; there was no equivalent gain available for
prose formatted this specifically, so those were updated by hand
instead, along with colitur-templates(5)'s FLAVOURS section, which
gained typst's own subsection alongside latex/groff/html/xml/ics/none.
test/cli.t's two pinned error-text assertions are updated to match,
and gained a new case: .typ inferred from extension, and a real
shipped citation's hyphen surviving --flavour typst's own escaping
live, the same shape the 2035 Ss. Petri et Pauli case already proves
for latex's ampersand just above it.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Adds templates/ef/ordo.typ (A5, one week per page, each day framed with
a colour swatch, real names, every fixed string through term.*) and
templates/ef/grid.typ (A4 landscape, one month per page, 7 columns,
colour-filled cells, Epistle/Gospel sigla), matching what ordo.tex and
grid.tex already do.
Typst resolves its own table of contents inside a single compile: no
pageref/aux-file dance, unlike pdflatex's own two-pass need for
ordo.tex. ordo.typ's TOC is one flat heading per week (Month . Week N)
rather than ordo.tex's own month-once/weeks-nested layout, since the
view model has no first-of-month flag for a template to test and
inventing one would be new engine state for a cosmetic grouping.
grid.typ deliberately diverges from grid.tex's own partial-fill
behaviour for a 5-week month: Typst's row-size list repeats its own
last entry for every remaining row, so a two-entry rows spec fills the
page completely regardless of week count, where LaTeX's fixed cellh
leaves a 5-week month's bottom sixth blank on purpose. Both divergences
are documented in the templates' own headers.
Two bugs found and fixed while building these against the real typst
0.14.2 binary: two stray hash characters in grid.typ's header comment
that were not actually inside a line-comment (would have been parsed
as Typst code, not prose); and two lines merging into one wherever they
were separated only by a bare newline (still the same soft-wrapped
paragraph in Typst, unlike LaTeX's own explicit paragraph break) --
fixed with a trailing backslash to force a real line break without
adding inter-paragraph spacing, verified against a full-year render
(zero typst warnings across all 365 days of 2027, in both templates).
Goldens (test/golden/ordo-2027.typ, test/golden/grid-2027.typ) were
generated through the harness's own View.of_days/Template.render_string
path, via a throwaway generator, NOT through the CLI: the CLI's default
language table differs from the harness's own English-chained-to-Latin
table, confirmed live (diffing CLI output against the harness's own
render showed exactly the expected language divergence and nothing
else) -- the same trap this project has hit before.
make check-templates now typesets both templates with the typst binary,
using the same SKIPPED-and-exit-0 shape the pdflatex/groff blocks
already use when their own tool is absent, verified in both directions:
absent (PATH without typst) skips loudly and exits 0, and a
deliberately corrupted template fails the target with a non-zero exit
and typst's own error text.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Adds Escape.Typst: type, all, to_string ("typst"), of_string,
of_extension (.typ), and a per-character expand function, structured
identically to the existing latex escaper (one pass, no re-scan, so
double-escaping stays impossible by construction).
The metacharacter set was verified against the installed typst 0.14.2
binary, not assumed: a probe document escaping each of #, *, _, $, @,
<, >, `, \, ~ and - was compiled and pdftotext'd back to confirm the
literal character survives, and each was separately confirmed to do
something else when left bare (# opens code mode, */_ toggle
strong/emph, $ opens math, @ opens a reference -- a bare unresolved
@word is a hard compile error, not merely mangled output -- </> can
close around a bare word into label syntax that swallows it whole, `
opens raw, ~ is a non-breaking space, and a run of two or three '-'
becomes an en/em dash). All ten are backslash-escapable; none needed a
non-backslash workaround. '-' is escaped unconditionally rather than
only inside a detected run, since this escaper has no lookahead -- a
probe confirmed escaping every hyphen independently still typesets as
literal hyphens for a run of any length, so the single per-character
rule is sufficient.
test_escape.ml's new Typst cases were written first and shown to fail
against a stubbed identity apply before the real escaper landed, per
this project's own regression-test discipline. Comments in both files
mark which new tests are genuine regression tests (the per-character
escaping, including a real shipped citation and a synthetic
markdown-habit overlay name) versus characterisation (to_string/
of_string/of_extension are flat table lookups with no logic to have
been wrong).
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
colitur-config.5 documents the file's location, the flag > config >
default precedence, every setting, and that an unknown key is warned
about rather than fatal. Its worked example is a real run against the
shipped example overlay, pasted rather than hand-typed -- including one
genuine gotcha found while verifying it: overlay/template values are not
shell-expanded, so a bare "~/..." written into the config file fails to
load; the full path must be written out instead.
colitur-templates.5's field reference is updated for the view change --
every field checked against lib/render/view.ml and schema/day-v1.json,
not transcribed from memory or an earlier version of this page -- and its
worked example is pasted from a real `colitur table` run. Its
scope-shadowing warning about {{name.la}} is REMOVED and replaced with a
short note that the hazard is now unrepresentable, because name is a
plain resolved string with no dotted path left to fall back through. The
num collision between a month and a week (both carry num, and a bare
{{num}} read from inside a day silently climbs to the week's own value,
not the month's) still exists and is demonstrated with a real render, not
merely asserted. A man page describing a hazard that was fixed is as
wrong as one that omits a real hazard.
Two further stale mentions of the removed {{name.la}} hazard, and two
places describing the schema as "Latin and English names" rather than one
resolved display name, were found in colitur.1 while cross-checking and
fixed alongside the primary rewrite; its own CSV example in EMIT was also
still the pre-Task-5 column list (name_la/name_en, no *_name columns) and
is now the real, verified header.
lang/ already installs beside templates/ and schema/ (a prior hotfix);
colitur-config.5 is added to install/uninstall/man/doc in the Makefile so
`make install`/`make doc` cover it too.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Names are Latin by default now; --raw restores the old byte-exact output,
verified against a pre-branch binary built from a clean HEAD checkout, not
merely asserted. day and readings gain a trailing display-name field
(day's own name suffix, readings' own " | name" suffix) rather than a
substitution in the slug's existing position: a name contains spaces and
would otherwise break the space-separated columns, the same reasoning that
already keeps readings a separate command. The field is present only when
it differs from the slug already shown, which is a direct consequence of
Lang.raw being the identity table (no lookup ever misses), not a special
case threaded through day_line/readings_line -- under --raw, or any
language with no entry for a given day, the trailing field is simply
absent, byte for byte what this project printed before naming existed.
An unknown language is an error naming what is available, never a silent
fallback to Latin: a booklet quietly printed in the wrong language is worse
than one that refuses to print. A declared fallback still chains, so a
partial translation shows its fallback language rather than slugs.
lang_dir mirrors data_dir's probe order exactly -- installed prefix, then
build tree -- because an installed binary that finds calendar data but no
language files is the failure this would otherwise ship.
Config (lang/overlay/template/format) is loaded once, unconditionally,
before dispatch: a config file that exists but fails to parse is fatal on
every command, not only the ones that consult it, because a config the
user wrote and colitur cannot honour is not something to silently carry on
past. An unknown key or unknown section is reported and ignored, worded
differently so a misspelled section reads differently from a misspelled
key inside a recognised one. overlay's config default applies only when
zero --overlay flags were given, never merged with a partial CLI list, so
precedence stays flag > config > default in one direction throughout.
colitur lang --list/--dump/--check and colitur config --show are what make
"anyone can write a language file" true rather than merely permitted, the
same pattern the overlay system already established with new-overlay and
check. --check reports what is missing AND rejects an unknown slug (exit
1), because a key matching nothing is silently dead and its author would
otherwise never learn why their own name does not appear. --dump emits
sections in a fixed order (Lang.keys' own global sort keeps every
section's keys contiguous), so two dumps of the same table are
byte-identical. config --show resolves through the same Config.resolve
every other command uses -- deliberately no separate provenance function,
so the two cannot disagree -- and previews exactly what flags given on the
same command line would resolve to anywhere else.
One authorized change outside this task's own file list: emit_ics.ml's
SUMMARY line was reading the unlocalised rank/colour pair ("class-1,
white") beside an already-localised name, where the JSON schema right next
to it correctly carries rank_name/colour_name ("II classis"/"albus").
Found while reviewing the published site; fixed at the source.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem 1: grid.tex's \daycell pushed the rank/class line to the exact
bottom edge of its fixed-height parbox with \vfill, the same edge \hline
draws the cell's bottom rule along, so 'I classis'/'IV classis'/'II
classis' printed as if sliced by the rule. Fixed by nesting a second,
shorter parbox (\cellh minus a new \cellpad, 1.5mm) inside the
unchanged \cellh-tall outer one, so the \vfill only ever reaches to
\cellpad above the rule, never touching it. \strut on both trailing
lines belt-and-suspenders the fix for lines with no descenders.
Problem 2: each cell now also carries the day's Epistle and Gospel
references (view fields first/gospel), joined by a centred dot/bullet
rather than repeated Ep./Ev. labels -- the same compact convention
ordo.tex already uses between rank_name and colour_name. A grid cell is
only ~3.5cm wide and Holy Week's own citations run 20-30 characters each,
so class label and sigla get their own \tiny line apiece rather than
being crammed onto one line and shrunk past readability; \cellh leaves
about eleven \tiny lines of room, so two short lines cost little. Both
first and gospel are wrapped in their own conditional section so a day
missing either prints no stray separator (defensive: every day in the
shipped EF lectionary has both, but nothing guarantees that of a future
rite or overlay).
grid.ms gains the same sigla, on their own .br line, using tbl's existing
w() wrap rather than any manual line-splitting -- it was never subject to
problem 1 (tbl auto-sizes row height, no fixed-height clipping risk) and
never carried a class label to begin with, so none was added here.
grid.html gains the sigla as a small, muted block-level span; unlike the
two paper flavours it isn't boxed onto a fixed physical page, so the
browser just grows the cell rather than needing any line-budget
compromise.
Verified with two pdflatex passes and groff -ms -t -Tpdf against all
twelve 2027 months: zero Overfull warnings, 7 columns every row, class
label and sigla both fully clear of every rule. test/golden/grid-2027.*
regenerated through test_render_golden.ml's own render path (View.of_days
+ Template.render_string, the EN-with-LA-fallback lang table), not the
CLI, per this project's own standing trap about the two tables differing.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
la.ini and en.ini (725 Latin/English liturgical names -- the entire
point of the naming-and-config work just merged) had no dune install
rule, so make install / dune install never copied them to
<prefix>/share/colitur/lang/. bin/main.ml's lang_dir() probes exactly
that installed path first, falling back to a build-tree layout that
also does not exist for an installed binary -- so every installed
user silently got raw slugs (ef-circumcision) and untranslated terms
(epistle/gospel) instead of resolved names, with no error at all.
The defect existed because nobody had run the installed binary from a
directory with no colitur source tree above it: dune exec and dune
test always resolve lang/ via the build tree, materialized as a side
effect of test/dune's cram deps, so every in-tree check stayed green
while the installed artifact was broken.
Add lang/dune, following schema/dune's exact pattern: an install
stanza placing la.ini and en.ini at <prefix>/share/colitur/lang/,
matching where bin/main.ml's lang_dir() already looked. No code
change was needed there -- only the missing install rule. Update
Makefile's install target help text and summary echo to mention lang
alongside ef/templates/schema.
Verified end to end: uninstalled, removed lang/dune, rebuilt and
reinstalled to reproduce the bug from an outside directory (~/tmp,
colitur table showed ef-circumcision/epistle); restored lang/dune,
reinstalled, and confirmed the same command now shows the resolved
Latin name and rubric terms, that ~/.local/share/colitur/lang/
contains both files, and that colitur emit --format csv shows a real
name in the name column. lang/la.ini and lang/en.ini are untouched.
Full suite: dune test --force, 495 tests, 0 failures.
|
| |\
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
colitur computed the calendar correctly and could not say what it had
computed. A printed ordo read ef-septuagesima-sunday-2 where a reader
expects Dominica in Sexagesima, and the wall calendar showed slugs in
every cell.
lib/naming a language table and a config file, both pure and total,
parsing the INI reader Overlay_ini already had
lang/ la.ini and en.ini -- 725 names, every one transcribed
from the 1962 Missal and citing the line it came from
templates the ordo rebuilt as an A5 booklet: one week per page, a
table of contents, framed days, a colour swatch; the wall
calendar now fills its sheet instead of a quarter of it
tools check_citations.py verifies all 400 citations resolve,
with 33 self-tests of its own
Names resolve through lang -> declared fallback -> the slug, so a partial
translation is usable from its first line and the fully degraded case is
the old output rather than a blank page.
colitur day and colitur readings are BYTE-IDENTICAL to before, verified
against main rather than asserted; --lang, --raw and the lang/config
subcommands are still to come, and man/colitur-templates.5 still
documents the pre-naming view, so writing a custom template needs the
source until that lands.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The six ordo booklet goldens (tex, ms, html, adoc, md, txt) were stale,
not broken: the commemoration-naming task landed 120 Latin names for
sanctoral slugs, so the render path test_render_golden.ml exercises now
prints real names (e.g. 'St. Telesphorus Pope and Martyr', or 'S. Felicis'
via en.ini's own la.ini fallback where English has no entry) where the
committed goldens still pinned the raw slug ('telesphorus-pope-and-martyr',
'felicis').
Regenerated through the same path test_render_golden.ml uses -- en.ini
chained to la.ini via Test_view.view_of, never the CLI's Latin-only
default -- with temporary env-gated scaffolding in the test file, used
once and reverted before this commit. Every changed line is a 1:1
slug-to-name substitution; file structure, line counts, and everything
else are untouched (570 insertions / 570 deletions across the six files).
Verified before committing: no unrendered {{ }} tags, none of the 332
known sanctoral/adjustment slugs leak into any of the six files, all
twelve month headings and all 365 day entries are present in each. Full
suite (495 tests), check-templates, and check-citations all still pass.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
test_lang_coverage.ml's own coverage test used to walk only the OBSERVED
day of each date (one Celebration.t per Liturgical_day.t). A liturgical
day also carries a second stream of slugs -- commemorations (kept when the
observed day does not fully displace a losing candidate, RG 108-111) and
transfers (an impeded I/II-class feast moved to a later date, RG 96-98) --
and nothing here ever looked at them. The test asserted coverage of what
it happened to WALK, not of what the engine can EMIT, so it passed green
while the ordo booklet printed raw slugs ("Commemoratio canute-martyr",
"Commemoratio maur-abbot", "Commemoratio peter"). test_lang_coverage.ml
now walks observed, every entry in commemorations, transferred_in, and
every entry in transferred_out.
That extension turned up 120 slugs with no Latin name across 2020-2045,
now added to lang/la.ini: 112 are data/ef/sanctoral.sexp companions the
temporal-only walk never reached; 5 (barbara, commemoration-of-st-peter,
commemoration-of-the-seven-sorrows, major-litanies, rogation-wednesday)
are data/ef/adjustments.sexp's own hand-authored Add directives; 3
(ef-nativity-octave-day-2/3/4) are temporal days with no calendarium
heading to transcribe, built by the same pattern days 5-7 already use.
Seven of the sanctoral names are cited to docs/research/scan1.txt rather
than LT.txt: the 2006 web-capture transcription silently drops several
real commemorations that both photographic scans carry (donatus, romanus,
eusebius-confessor, naboris-et-felicis, sts-gervasius-and-protasius,
sts-felix-and-adauctus, and praxedis-virginis -- the last for a spurious
ligature LT.txt introduces, "Praxedis" confirmed correct against both
scans). The same 120 names, extracted verbatim from sanctoral.sexp's and
adjustments.sexp's own English fields rather than retyped, are added to
lang/en.ini.
make check-citations: 400 LT.txt citations checked, 0 wrong, 0 malformed,
0 cannot verify. Teeth proved by deleting canute-martyr's own name (a
Commemoration_only slug that can never be the observed day, only ever a
commemoration) and confirming the coverage test fails naming exactly that
slug, then restoring it.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The previous version occupied about the top quarter of an A4 page and
left the rest blank, and printed slugs. Cell height is now computed
from the text height (\textheight-18mm over 6 rows -- every month has
5 or 6 Sunday-started weeks) rather than left to the table's natural
size, which is what made it float at the top. Measured: January 2027's
last row of content now reaches y=566pt of a 595pt-tall page, not a
quarter of the way down.
Real names replace the raw slug in every cell (the observed day's own
resolved display name, a plain string per View.of_days), and each
day's rank name is shown too, so a printed page reads as a calendar
rather than machine keys.
Weekday headings come from the view as a localised list rather than a
hard-coded Dom/Lun/Mar row, so a translated calendar needs no template
edit -- and because the engine rejects an empty tag path, the list
carries named {name; last} fields rather than bare values.
grid.ms and grid.html carry the same two changes (real names, localised
weekday headings) for their own flavours.
Verified with pdflatex/groff: 12 pages, A4 landscape, every row exactly
7 cells (6 ampersands), zero warnings. Goldens regenerated through the
test harness's own render path, not the CLI, whose default language
table differs from the test's.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
A LaTeX table of contents needs two pdflatex passes -- the first pass
leaves every entry showing '??', the second resolves the \pageref
values. A user hit exactly this and reported it as a bug; the template
was fine, the instructions were not.
README's rendering example now runs pdflatex twice for ordo.tex (or
names latexmk -pdf as the one-shot alternative) and notes the wall
calendar needs only one pass, having no cross-references of its own.
make check-templates now runs pdflatex twice per LaTeX template too,
so the target exercises what a user actually has to do rather than
silently passing on a single, incomplete pass.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
geometry's top=11mm/bottom=12mm seated the text block without accounting
for fancyhdr's own header/footer, which live OUTSIDE that block by
default -- pushing the running header ('Ordo 2027 . ef') partly above
the physical page edge (measured: yMin -5.5pt on a 0..595pt page) and
the footer to within 1pt of the bottom edge. The user hit this and
reported the header as clipped.
Adds includehead/includefoot to the geometry options (a5paper,
top=10mm, bottom=10mm) so the header/footer live inside the margins
instead, with headsep/footskip set explicitly (3mm/7mm) rather than
left at article's defaults -- those defaults alone would still overflow
a 10mm margin and silently added 20 extra pages to the whole booklet by
shrinking every day box.
Verified: two pdflatex passes produce a 65-page A5 PDF with the full
'Ordo 2027 . ef' header intact on every page and zero pdflatex
warnings. Golden regenerated through the test harness's own render
path (English-with-Latin-fallback), not the CLI, which defaults to
Latin only and would otherwise pin output the tests never produce.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
A5, one week per page, with a table of contents, each day in a framed box
and a colour swatch so the page is scannable without reading. Real names
throughout, and every fixed string routed through the term vocabulary so
a translated booklet needs no template edit.
The previous version was A4, printed slugs, ran days together with no
visual separation, had no contents, and gave the liturgical colour only
as the word white. It was not usable as a printed booklet.
lib/render/view.ml already carried month_num/month_name on each week
object from Task 5, so no kernel/view change was needed here -- the
template uses those instead of a parent path the engine cannot express.
Box padding, margins and secondary-line font size were tightened past
the brief's own starting values: at the brief's sizes a week with several
long commemoration names (e.g. Feb 21-27, three of seven days carrying
one) overflowed onto a second physical page, breaking one-week-per-page.
Verified by measuring page count against the known week count (63 weeks
in 2027) until every content page carried exactly one week, not by eye.
The header comments of all six flavours were themselves a trap the brief
warns about: writing double braces to NAME a template field inside a
LaTeX %, groff .\", HTML <!-- -->, or AsciiDoc // comment gets parsed as
a real tag by this brace-only engine, not treated as inert prose. An
early draft's own comments did this and produced 'empty tag path' parse
errors and a silently-unclosed section; every comment in all six
templates is now written without ever typing two braces in a row.
The other five flavours keep their existing structure; only the dead
{{#name}}{{la}}{{^la}}{{slug}}{{/la}}{{/name}} idiom is replaced with a
plain {{name}} (extended to comms entries too, which carry their own
resolved name), and the fixed labels (Ordo, Epistle, Gospel,
Commemoration) now come from {{term.*}}.
Golden regeneration could not follow the brief's own `colitur table`
shell-out literally: the CLI's current default language is Latin-only
(bin/main.ml's Task-5 bridge), while test_render_golden.ml renders
through Test_view's English-primary/Latin-fallback table, so the two
produce different text for the same slug. The goldens were regenerated
through the test's own render path instead (a temporary env-gated block
in test_render_golden.ml, reverted before this commit), so they agree
with what the suite actually computes.
All six ordo golden tests pass; the three grid ones are Task 9's scope.
make check-templates passes with zero warnings. The CLI-rendered PDF is
A5 (148x210mm), 65 pages -- 2 pages of title/contents plus exactly one
page per week (63), verified with no overflow.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Ran the round-2 hardened check_citations.py against lang/la.ini and fixed
every WRONG and MALFORMED finding. Baseline before this task: 12 WRONG,
42 CANNOT VERIFY (round-1 tool). After the round-2 tool alone (data
unchanged): 9 WRONG, 10 MALFORMED, 1 CANNOT VERIFY -- the rarity rule
cleared 41 of the 42 CANNOT VERIFY entries by itself, as expected. After
this cleanup: 0 WRONG, 0 MALFORMED, 0 CANNOT VERIFY, 292 citations checked
(up from 272 -- several wide ranges became more granular, individually
verified citations).
What changed, all in comments, no `key = value` entry touched:
- Three genuine wrong citations found and corrected: advent's own
citation was pointing at christmastide's real heading (LT.txt:8631,
"Tempus Nativitatis") 22 lines off from its own (LT.txt:8609) --
christmastide had never had a working citation of its own at all, the
right line sitting unused inside advent's corrective note. Restructured
so each entry cites its own real line, including giving
time-after-epiphany its own citation (LT.txt:8668) for the first time.
- Ten MALFORMED wide ranges (introduced by round 1's own tolerance
removal, never capped until round 2) replaced with precise per-entry
citations -- mostly comma lists of exact bare line numbers, since each
TOC section lists one heading per line; two are legitimate multi-line
wraps kept as capped explicit ranges (a 2-line Nativity-octave heading
split by an unrelated saint's day; Christ the King's own in-body heading
split across a page-number line).
- Six citation-shaped substrings that were never genuine provenance
claims lost their "LT.txt:" prefix (now read "line NNNN"), each with an
inline note explaining why: two were corrective prose quoting a
historical WRONG value ("previously cited LT.txt:8631/12459, which
is..."), three were contextual pointers to a NEARBY but different
heading used for explanation, and one discloses an unmodelled alternate
wording. None of these ever claimed to be this entry's own heading;
writing them as "LT.txt:N" only let the checker mistake documentation
for a claim.
- Two genuinely correct citations left permanently unable to pass an
automated word-overlap check, for reasons orthogonal to correctness
(disclosed in check_citations.py's own docstring as an accepted
trade-off) also lost their "LT.txt:" prefix, each verified by hand and
noted as corroborating rather than primary: class-1..4's "classis" (the
only distinctive word in "I classis" etc., 500+ occurrences across
LT.txt -- RG 8 is the primary source); Corpus Christi's and Holy Name's
own TOC lines, both abbreviated ("Ss.mi"/"Ss.mae") where every entry
below spells the same title out in full, and neither has a spelled-out
occurrence anywhere else in this partial 2006 web-capture transcription
to cite instead (Holy Name also has a stronger primary source already:
temporal_ef.ml's own scan-verified string).
No rule was weakened to reach zero: every de-prefixed reference was
verified by hand against the transcription first, and none of them was
ever wrong -- each was either documentation, context, or evidence the
checker's own disclosed word-overlap/rarity design cannot confirm.
Claude-Session: https://claude.ai/code/session_017ZBxCCRM2ojnBupp3SBxV9
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Round 1 hardened check_citations.py against its own self-poisoning bug;
a review defeated it again. Five fixes, in the order they were found:
1. PATTERN silenced a whole comment block, not just the entry it was
attached to -- a wrong citation on a DIFFERENT, unmarked entry in the
same block (e.g. [season]'s own back-to-back trailing-comment style)
was never checked at all. Fixed by scoping PATTERN with the identical
leading/trailing pooling rule citations already use: an entry is
excluded only by its own marker, never a neighbour's.
2. Explicit per-citation ranges (introduced in round 1 to replace a
blanket +-2-line tolerance) had no upper bound, reintroducing the same
defect at a much larger radius (LT.txt:8600-8650 passed if the text
appeared anywhere in fifty lines). Capped at MAX_RANGE_WIDTH (3 lines);
anything wider is reported MALFORMED, naming the entry and the width,
instead of silently accepted.
3. The "pool too thin to verify" gate counted words, not rarity -- it
flagged 11 genuinely correct citations (short Latin hagionyms with only
one non-stopword) CANNOT VERIFY, while a match on nothing but "classis"
(507 occurrences) passed freely alongside three siblings. Replaced with
a frequency table over the whole LT.txt corpus: a token's evidence is
1/(times seen), an item's evidence is its single rarest matched token
(not a sum -- summing would let several merely-common words add up to
"enough" between them, the same shape as the self-poisoning bug).
4. "LT.txt:12,459" (a comma typo for one number) parsed as two unrelated
bare citations, 12 and 459, either of which could coincidentally match
while the intended line was never checked. Detected as a
thousands-separator-typo shape (a 1-2 digit token immediately followed
by an exactly-3-digit one -- the only way a real LT.txt line number,
which never exceeds 5 digits, splits under one comma) and rejected as
malformed.
5. The self-test suite overstated its own coverage: of round 1's seven
fixture cases, only two actually failed against the pre-round-1 script.
Every test is now labelled REGRESSION or CHARACTERISATION, each
verified by direct replay against the named prior version rather than
asserted -- 14 of 33 are genuine regression tests.
Both of the review's own defeats (block-wide PATTERN silencing, the
50-line range) are reproduced as dedicated fixtures and confirmed caught;
both are also confirmed to slip through the pre-round-2 tool unchanged.
Claude-Session: https://claude.ai/code/session_017ZBxCCRM2ojnBupp3SBxV9
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The view's name is now the RESOLVED display string and slug is untouched,
so machine formats carry both -- a script keeps the stable key, a human
reads the name.
name is a plain string, not a lang-keyed object. That removes the
shadowing hazard outright: a dotted {{name.la}} used to fall back
WHOLESALE to the enclosing month's name.la and print Ianuarius on every
unnamed day, which is how a printed booklet came to show the month where
the feast belonged.
weekday, season, rank and colour all gain localised companions, because a
calendar in a language needs more than feast names, and templates gain a
term vocabulary so fixed strings need no template edit to translate.
Asserted over a whole year: no day renders its slug as its name.
Beyond the brief's own code sample:
- comm_value's own name is now resolved through the same lang.celebration
table too (not only the observed day's), because Task 8's own ordo
template interpolates a plain {{name}} inside {{#comms}} -- an Obj there
would render silently blank. A commemoration slug without Latin
coverage still degrades to the slug, same as everywhere else in this
system; that is a lang/la.ini DATA gap (113 of 327 sanctoral slugs,
measured), not a regression this task introduced.
- bin/main.ml's emit/table/publish call sites needed ~lang to compile at
all, which is collateral from the of_days signature change, not this
task's own file list. Rather than pass Lang.raw and ship the very
slug-as-name defect this branch exists to fix, they load the shipped
Latin table by the same probe order data_dir() already uses -- a
deliberate, commented BRIDGE that Task 6 replaces wholesale with real
--lang/--raw/config resolution. bin/dune gained colitur_naming
accordingly.
- test/cli.t needed two related fixes to stay green: the CSV header/row
example, and a table/LaTeX escaping demonstration that relied on the
kernel's own English name for Sts Peter & Paul -- gone from the view
now that name resolves through lang tables only, and the Missal's own
Latin spells the feast with et, never an ampersand. Escaping itself is
still proved live on 2035 data in test_emit.ml.
- Both schemas gained the new day/week/top-level keys (season_name,
weekday, rank_name, colour_name, term, weekday_headings, month_num,
month_name), not only the name shape change; schema/colitur-v1.xsd
verified against real emitted XML via xmllint (make check-schema).
Render/golden's 9 cases (the shipped ordo/grid templates, all six
flavours) now fail as expected: their old {{name.la}} / {{#name}}...
idiom finds nothing on a plain string. That is Tasks 8/9's own scope to
rewrite, per the plan's own pre-flight conflict scan -- not fixed here,
and not silently pinned by regenerating goldens off broken output.
495 tests run (490 + 5 new), 486 pass; the 9 failures are exactly
Render/golden's ordo/grid cases.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Sanctoral names are transcribed from the Missal's own calendarium and
kept in the GENITIVE, as the Missal prints them -- noted in the file so
nobody corrects them to the nominative.
English reuses the 327 names already in data/ef/sanctoral.sexp rather
than retyping them, and declares fallback = la, so an untranslated day in
an English booklet shows Latin rather than a slug. The test asserts the
FALLBACK works rather than that en.ini is exhaustive: that is what makes
a partial translation shippable from its first line.
Coverage now demands a name for every slug the engine can emit across
2020-2045, temporal and sanctoral alike.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Reproduced the defect: reintroducing the exact historical citation bug
(pointing class-1's citation back at LT.txt:12459, the value a prior
fix round corrected away from) made the tool report "147 citations
checked, 0 look wrong". The mechanism was that the corrective comment
documenting the old bug quotes the wrong historical value, and the
checker pooled every quoted phrase from the whole surrounding comment
block, so citing the wrong line matched the comment explaining why it
was wrong.
Four changes:
1. The word pool for a citation is now scoped to the entry(ies) it is
attached to only -- never to quoted text elsewhere in the comment.
This is the direct fix for the self-poisoning bug.
2. A citation whose pool has fewer than two distinctive words (Latin
liturgical headings are short and stopword-heavy) cannot
discriminate the right line from a wrong nearby one. Such a
citation is now reported CANNOT VERIFY and fails the target,
instead of silently passing.
3. The blanket +-2-line tolerance is gone. A bare "LT.txt:N" is
checked at line N only; a heading that genuinely wraps must say so
explicitly as "LT.txt:N-M". The allowance moves into the data,
where it is visible.
4. The tool gets its own test suite, tools/test_check_citations.py,
with a synthetic fixture covering: a correct citation, off-by-one
and off-by-three mismatches, an explicit wrap range, a degenerate
pool, a PATTERN-marked entry with no citation, and a dedicated
regression test for the self-poisoning case itself. Wired into
`dune test` via a new (rule (alias runtest) ...) in tools/dune (a
plain (test ...) stanza cannot run a Python script), so it runs
with the rest of the suite, not only as a `make` target.
Added a --file/--lt-file override to check_citations.py so the tool
(and its own tests) can point at a fixture without touching the real
lang/la.ini or docs/research/LT.txt. Confirmed the "SKIPPED, exit 0"
behaviour for a missing docs/research/LT.txt is unchanged.
tools/__pycache__/ (a stray artefact of this script, previously
untracked and ungitignored) is now in .gitignore.
Measured against the current lang/la.ini (another task is still
landing its sanctoral entries on this branch): 15 of 275 citations now
look wrong and 42 more cannot be verified, both far above the 0 the
unhardened tool reported. Not fixed here -- the data pass is separate,
once the sanctoral entries land.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Two of la.ini's LT.txt:<n> citations pointed at the wrong line -- the
Latin itself was right, only the pinned line was wrong:
- advent cited LT.txt:8631 ("Tempus Nativitatis"); the real "Tempus
Adventus" heading is at 8609.
- ef-christ-the-king and [rank]'s own citation both pointed near
"Dominica ultima Octobris" (12459) when the text they actually quote,
"D.NI NOSTRI JESU CHRISTI REGIS" and "I classis", sits two and three
lines further down, at 12461 and 12462.
ef-christmas-sunday-0 was marked PATTERN but LT.txt:8644 is the identical
string verbatim -- relabelled as a direct citation, not constructed.
Added tools/check_citations.py and `make check-citations`: for every
LT.txt:<n> citation outside a PATTERN block, confirms a +-2-line window
around line n actually contains the Latin text the citation claims,
rather than trusting each of the 38 citations by hand. Follows
check-schema/check-templates' own precedent -- docs/ is gitignored, so
the target prints SKIPPED loudly and exits 0 when docs/research/LT.txt
is absent, never a silent pass.
The checker's own teeth are proven three ways: replayed against the
pre-fix file it independently re-derives both corrections above; a fresh
mutation (redirecting one citation to an unrelated line) is caught and
reverted; the fixed file passes clean, 147 citations checked, 0 wrong.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Every entry is transcribed from the 1962 Missal's own propers headings in
docs/research/LT.txt and cites where it came from; names constructed by
following a neighbouring pattern are marked as such, so a reader can tell
transcription from inference.
The coverage test is the point of this commit. It walks every day of
2020-2045 and fails naming any slug with no Latin name -- the test that
would have caught the original defect, where a printed booklet said
ef-septuagesima-sunday-2 because nothing asserted that names exist.
The three Triduum names reuse the exact strings temporal_ef.ml already
carries, so the engine and the language file cannot disagree.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
config.ml and lang.ml both parse the INI format through the same reader,
Colitur_kernel.Overlay_ini.parse_sections, but resolved a repeated
[section] header oppositely: lang.ml folds over every section sharing
a name, while config.ml used List.find_opt and silently discarded every
[defaults] block after the first. Two modules parsing one file format
must not disagree about what a duplicate section header means.
of_string now folds a single accumulator across every section named
[defaults], in file order, matching lang.ml's of_string shape. A
scalar key (lang/template/format) repeated across two blocks resolves
to the later value, consistent with the existing within-section
last-wins rule; overlay keeps accumulating across every block, not
only the first; and unknown_sections still excludes every [defaults]
block, merged or not, since merging it is the point.
config.mli's lang doc comment is extended to say the last-wins rule
holds across block boundaries too, cross-referencing lang.ml's own
duplicate-section policy so the two do not drift again unnoticed.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
F1: test_unknown_key_is_reported_not_fatal never asserted unknown_keys
itself, only that parsing survives -- a no-op accumulator passed it.
Now asserts the key is actually collected.
F2: a misspelled section name, e.g. [deafults], was silently discarded --
Ok empty, lang and everything else gone, nothing reported. That is the
highest-value typo this feature exists to catch. Any section other than
[defaults] is now collected into a new Config.unknown_sections, kept
separate from unknown_keys so the CLI can word the two warnings
differently. Still non-fatal: a newer colitur's added section must not
break an older binary.
F3: overlays and unknown_keys accumulated with '@ [v]' per line, O(n^2)
over the field count. Cons during the fold, List.rev once at the end.
F4: documented that lang/template/format are last-wins on a repeated
key, the opposite direction from Overlay_ini.get's first-wins over the
same section type.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
F1 (review round 1): of_string's find took only the FIRST section of a
given name (List.find_opt), so a second [celebration] block anywhere in
the file was silently dropped in its entirety -- reproduced with two
blocks (a in the first, b in the second): b resolved to the slug
fallback "b", not its real value.
This is a data-loss footgun aimed squarely at what happens next: Tasks
3/4 write a 595-entry, hand-edited la.ini, and appending a second
[celebration] block is the natural way to paste in a new batch of names.
Worse, the failure surfaces nowhere near its cause -- a coverage check
reports the dropped slugs as missing a Latin name, with nothing pointing
at the parser.
find now folds over every section sharing the name, in file order, so
all blocks merge. This also settles which value wins when the same key
appears in two different blocks: later in the file wins, consistent with
the existing within-one-block behaviour (unchanged, still last SM.add
wins) and with what a reader expects when appending to an INI file.
lang.mli now documents both duplicate policies explicitly, and notes
they run OPPOSITE to Overlay_ini.get's first-match (List.assoc_opt) over
the same section.fields shape -- undocumented before, and a latent trap
since the two modules read the same section type but resolve a
duplicate key in opposite directions.
Three tests added: two [celebration] blocks both resolve (the F1
regression), a key repeated across two blocks resolves to the later
block, and a key repeated within one block still resolves to the later
line (confirms unchanged behaviour). Confirmed the regression test fails
against the pre-fix code (b resolves to "b", the slug fallback) and
passes after.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Owns precedence and provenance and nothing else, and never reads the
filesystem, so it is as testable as the language table.
resolve returns the value AND its source, because a setting that silently
comes from a file the user forgot about is worse than no setting at all --
config --show can then say where each effective value came from.
overlay accumulates rather than last-wins: a user has more than one.
An unknown key is reported, never fatal. A config written for a newer
colitur must still work on an older one, but silently dropping a line the
user wrote is how a typo becomes invisible.
|
| |/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Maps strings to strings and nothing else -- no calendars, no dates, no
filesystem. That is what lets every command use it without the kernel
learning about presentation.
Every lookup is total, and a miss returns THE KEY rather than the empty
string. A partial translation is therefore usable from its first line,
and the fully-degraded case is exactly today's output (bare slugs) rather
than a blank page.
--raw is a real identity table, not a special case threaded through every
call site: one value the whole program passes around.
Reuses Overlay_ini's INI reader rather than growing a second one that
would drift in its comment, quoting and trimming rules; parse_sections is
exposed in the .mli for that, with no behaviour change.
Fixes one defect found while running the brief's own tests rather than
transcribing them blind: weekday's internal lookup key is an English
day-name word (month's is already the numeral string), so on a miss it
echoed that word instead of the documented numeral, breaking both the
0=Sunday convention and Lang.raw's own identity contract for weekday.
weekday/month now fall back to string_of_int n directly on a miss
instead of through get's generic echo-the-search-key path; month is
byte-identical since its key already equals string_of_int n.
|
| |\
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Gives colitur a publishable exit. Until now its only output was terminal
rows; it can now print an ordo booklet and a wall calendar, publish an
iCalendar feed people subscribe to, and serve a static JSON/XML API.
lib/render escaping (six flavours + RFC 5545 folding), a deliberately
logic-less template engine, the view model, and five
emitters (CSV, JSON, XML, iCalendar, S-expression)
CLI emit, table, render, publish -- all accepting --overlay
templates ordo booklet in six flavours, wall grid in three
schema day-v1.json and colitur-v1.xsd, the published contract
man colitur-templates.5, plus colitur.1 updates
The view model is why the engine can stay logic-less: a month grid needs
leading blank cells, week bucketing and an in-month test, and a
logic-less template can compute none of it. Shaping the data in OCaml
keeps the engine safe for untrusted templates and makes the grid
trivial.
Formats split by whether correctness is mechanical. Presentation goes
through templates; iCalendar and XML get dedicated emitters, because
folding, exclusive DTEND, stable UIDs and schema fidelity are rules a
template cannot enforce and each fails silently in a subscriber's
client rather than loudly at generation.
publish is deterministic and non-destructive: two runs produce a
byte-identical tree, and --prune removes only files a previous run
created, refusing any manifest entry that escapes the output directory.
No new dependencies. The kernel and rite modules are untouched, and
colitur day and colitur readings remain byte-identical.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
It read "an inner key silently loses to an outer key of the SAME NAME" --
false: the document's own num entry, two paragraphs below, shows the
opposite (inside {{#weeks}}, a bare {{num}} is the week's own number,
correctly, because the inner scope wins). The real hazard, matching the
body text and the name example that were already correct: a DOTTED path
that resolves only part way inward (the day's own name object exists but
lacks the key the path needs) falls back WHOLESALE to an outer scope of
the same name, not a bare key losing outright. Corrected the headline
only; re-linted clean with groff -man -Tutf8 -ww -z.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
publish's own mkdir_p/write_file (unlike every other IO path on this
branch) were unguarded: an unwritable --out parent raised a bare
Unix.Unix_error(EACCES,...) and an --out naming an existing file raised
ENOTDIR, both as uncaught exceptions with a stack trace rather than the
project's one-line "colitur: ..." form. The same defect class commit
6bd741b already fixed once for template reads -- --out is user input
too. Fixed by wrapping the whole publish_report call (not each
write_file site) in one handler for Unix.Unix_error and Sys_error,
mirroring why that earlier fix guarded the whole read and not only the
open. Added a cram case using a read-only directory inside the test's
own cram sandbox, not /tmp, so a failed cleanup cannot leave an
unwritable directory behind in a shared location.
--dtstamp was the only user string reaching output unescaped and
unvalidated: "--dtstamp hello" silently emitted an invalid
"DTSTAMP:hello", and a value carrying its own CRLF injected extra lines
into every VEVENT. Fixed by rejecting anything not matching RFC 5545's
UTC form (8 digits, 'T', 6 digits, 'Z') before either emit or publish
does anything else, one line to stderr, exit 2.
usage() was byte-unchanged from before the branch and listed only the
six pre-existing commands, omitting all four commands this branch added
(emit, table, render, publish). Added them; the three cram pins of the
exact usage string are updated to match.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
pdftotext -layout of the shipped grid.ms showed only 3-4 of 7 columns and
roughly a third of each month's day numbers: tbl's plain columns (no
w()) size to their widest single entry and never wrap, so a long
fallback slug (some run past 40 characters) forced every column that
wide, the table ran far past the page, and whatever fell past the
physical edge was gone, not merely ugly. groff exited 0 throughout
(warnings, not errors), so make check-templates reported OK on a broken
artefact.
Fixed both halves. (a) The table now fits: true landscape via gropdf's
own -P-pa4l (an in-document Xpapersize=a4l escape was tried and
rejected -- it does not rotate the page in this groff), ms's own title
macro widened back out after narrowing the line length for the title
text (a second, independent way the original lost its width, found by
reading s.tmac), and every column rewritten as a genuine tbl text-block
(T{/T}, not a plain w() cell -- w() alone does not wrap, confirmed
against tbl's own generated troff code) so long, hyphenated slugs wrap
at their own hyphens instead of forcing the column wider. (b)
check-templates now captures groff's stderr per template and fails the
target if it is non-empty, rather than trusting groff's exit code.
Verified: 0 warnings (was 12), pdftotext -layout shows all 7 columns and
every day number for all 12 months (was 3-4 columns, ~12-23 of each
month's day numbers). The golden fixture is regenerated: 0 "{{", 12
month headings, exactly one Ianuarius, and each week's block (now spread
across several physical lines by the T{/T} wrap) carries exactly 6 tabs
joining its 7 cells.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The live-data CSV assertion inspected only the first field's length,
always the 10-char ISO date, which can never contain a comma -- it could
not fail no matter what emit_csv.ml did with the rest of the row.
Mutation-proved: replacing emit_csv.ml's escape_field call with identity
left every test green while the real output emitted a 14-field row
against a 13-column header. Fixed by parsing the row as RFC 4180 actually
requires (a small quote-aware splitter) and asserting the field count
matches the header, plus asserting the quoted substring appears literally.
The XML suite asserted Escape.Xml's correctness in isolation but never
that emit_xml.ml actually calls it on every interpolated value. Bypassing
one escape call at the name-element site left all tests green while real
2035 output (Sts. Fabian & Sebastian, 20 January) emitted a bare '&' that
xmllint rejects. Fixed by adding a live-data test against the 2035 fixture
asserting an escaped ampersand is present and no bare one remains.
Both new assertions were run against their named mutations and confirmed
to redden before being reverted.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The worked minimal template iterated the top-level, flat days list
and claimed the naive {{name.la}} form would print the enclosing
month's name there. It would not: with no {{#months}} wrapping it,
the flat days list has no month anywhere on the scope stack, so the
naive form on an unnamed day resolves to nothing, not to the month.
Reproduced against the live engine before touching the page: the
flat shape renders empty; the identical naive form nested inside
{{#months}} genuinely does print the month's name on both unnamed
days. The example now nests days inside months -- the shape every
shipped template actually uses, and the shape the hazard needs to
fire -- with a second, verified rendering showing the collision for
real, and a closing note stating plainly that the flat shape does not
reproduce it. SCOPE AND LOOKUP's own name-collision paragraph gained
the same nesting precondition it was missing.
Both rendered blocks were checked mechanically: extracted verbatim
from the page source and diffed against a fresh colitur table run
over the exact templates shown, byte for byte.
One cosmetic fix along the way: a transition sentence embedded the
full safe-idiom string inline via .B, which groff's justifier
stretched into visibly wide gaps when rendered. Reworded to reference
the idiom shown above instead of repeating it.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
CRITICAL: .colitur-manifest lives INSIDE the tree publish writes into --
the very tree this feature exists to have committed into a git repo. A
manifest entry with a ".." path component, or an absolute path, let
--prune Sys.remove/Unix.rmdir a file OUTSIDE --out. No attacker is
required: an ordinary bad merge, a conflict resolved the wrong way, or
a hand-edit of that file is enough to plant such an entry, and publish's
own stated contract -- it never deletes a file it does not own -- broke
outright the moment one was present.
Two independent checks, both required, applied before every deletion:
- structural (manifest_entry_is_safe): reject an entry that is
absolute or has a ".." path COMPONENT, by splitting on '/' and
comparing components, not by substring-matching ".." (which would
wrongly reject a legitimate name like foo..bar).
- containment (resolves_under): resolve both --out and the candidate
with Unix.realpath (closing a symlink-inside-out gap the structural
check alone would miss) and verify the candidate is a genuine path
descendant of --out, not merely a string with the same prefix.
Applied at both the file-deletion loop and prune_empty_dirs' own
directory removals. A rejected entry is skipped with a one-line stderr
warning; publish completes rather than aborting -- a corrupted manifest
must not make the tool itself unusable.
test/cli.t reproduces the exact canary scenario (a ".." entry
surviving deletion of a file outside --out), an absolute-path entry,
and a legitimate dotted filename (no .. component) still pruning
normally, alongside the existing --prune coverage.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
colitur-templates.5 documents the four syntax forms, the six flavours
and their escaping, and the full view-model field reference. It states
plainly that there are no partials, no raw form and no expression
evaluation -- a template is data, never a program. It documents two
real hazards found during this build, not theoretical ones: the
outward scope fallback silently shadowing an inner name/num key with
an outer one of the same name (with the safe {{#name}}...{{^la}}
idiom), and the engine's lack of host-comment awareness (a {{...}}
inside a LaTeX %, groff .\" or HTML <!-- --> comment is still parsed
as a tag).
It also states the limitation rather than hiding it: AsciiDoc and
Markdown are not escaped, so a feast name containing * or _ renders as
emphasis.
templates/ and schema/ now install into <prefix>/share/colitur/,
matching data/ef/, via new install stanzas; colitur-templates.5
installs to man5 beside colitur-overlay.5. Verified against a scratch
prefix: the installed binary resolves both from the prefix, not the
source tree, when run from an unrelated working directory.
make check-templates typesets every shipped template through pdflatex
and groff when they are installed, and prints SKIPPED loudly when they
are not. Golden tests prove templates render; only this proves they
typeset. A silent skip would read as a pass.
Fixed a real doc/help drift while here: bin/main.ml's --help still said
--overlay was accepted on day and readings only, three commands out of
date (emit, table/render and publish all accept it too), disagreeing
with the man page's own OVERLAYS section, which carried the identical
stale line. Both are corrected; --overlay's own behaviour is unchanged.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Writes ef/<year>.{json,csv,xml,ics}, one JSON per day, the schema and a
generated index. That tree is the API: any web server or git repo serves
it, and nothing runs at request time.
Deterministic: publishing twice is byte-identical, asserted in cli.t.
That is what makes publishing into a git repo safe -- the diff shows
only real change, and you review it before pushing.
Non-destructive: a manifest records exactly the files this tool wrote,
so --prune can only remove files a previous run created. A file you put
in the output directory yourself is never touched, with or without
--prune. Asserted in both directions.
Pruning a stale file also removes any directory it leaves empty behind
it (e.g. an old year's own ef/<year>/ tree), stopping at --out itself --
without this, a pruned year's own directory would survive empty and
test -d would still see it.
schema/day-v1.json is resolved the same prefix-relative way data/ef's
own sexp files are (installed vs build-tree, probed rather than
assumed), never from cwd, and a missing schema fails with one line on
stderr before anything is written rather than emitting an empty file.
Needed schema/day-v1.json wired into the root dune file's default alias
and into test/dune's cram deps -- unlike data/ and templates/, nothing
made dune mirror schema/ into the build tree before this.
unix is added to bin/dune's libraries for mkdir_p; it ships with the
compiler, so colitur.opam and dune-project are unchanged.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Three flavours, not six: a month grid in Markdown or plain text is a
worse artefact than the booklet already is, and shipping a template we
would not use ourselves is maintenance with no reader.
These are the first templates to use weeks and in_month, so this is
where the view model earns its keep -- the booklet and the grid come
from one model with no second code path.
Every cell carries a per-cell 'last' boolean (already in the view). A
table row needs a separator BETWEEN cells and the engine deliberately
has no 'unless last' construct; the rule is shape the data, not the
template. Without it the LaTeX grid emits eight columns for seven
cells and pdflatex rejects the file.
Day cells resolve their label as {{#name}}{{la}}{{^la}}{{slug}}{{/la}}
{{/name}}, never a bare {{name.la}}: the enclosing month object has
its own name.la, and the engine's scope lookup falls back outward, so
a bare dotted lookup would render the month's own Latin name on every
day lacking one -- a wall calendar where every day reads "January".
Verified on the goldens: Ianuarius appears exactly once per file, the
month heading, never as a day label.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
read_file guarded open_in_bin but left in_channel_length and
really_input_string unguarded, so a path that opens but cannot be read
as bytes -- a directory -- escaped as an uncaught Sys_error and crashed
the program, leaking the open channel on every failure path. A template
is user input; it must never crash the program.
Wrap the whole read in Fun.protect so the channel closes on every path
(success, exception, early return), matching the close-on-every-path
pattern already used in the test suite. The missing-file message stays
exactly as before; a read failure after a successful open now carries
the exception text, the same path: exception shape Layer.load and
Overlay.load already use.
New cram case points --template at a directory (the sandbox's own cwd,
not /tmp) and asserts one stderr line and exit 2, not a crash.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
LaTeX and groff are the print paths; HTML carries a print stylesheet;
AsciiDoc, Markdown and plain text are the plain-consumer paths.
AsciiDoc and Markdown use flavour none, and say so in a comment: their
metacharacters are context-dependent and escaping them aggressively
produces worse output than not escaping. The consequence is real and
documented -- a feast name containing * renders as emphasis.
Golden tests pin all six byte-for-byte for 2027. They prove the
templates RENDER, not that they TYPESET; compiling needs TeX and groff,
which is Task 13's opt-in make check-templates.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Computes and renders in one process. There is deliberately no stdin-fed
render: honouring the pipe would need a JSON parser we would have to
write, purely to serialise and immediately re-parse our own view -- a
second hand-rolled component and a second place for the contract to
drift, for no benefit. colitur emit --format json | jq still composes.
An unknown extension with no --flavour is an error naming the six valid
flavours, never a silent fallback to none: guessing wrong produces
malformed output that looks fine until it does not.
A malformed template reports the parser's own reason and exits 2. A
template is user input; it must never crash the program.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
F1: Date.add_days is UNBOUNDED (date.mli) -- only Date.make enforces
1583..9999 -- and Date.to_iso8601 pads but never truncates, so
9999-12-31's naive successor formatted as "10000-01-01", and compact
turned that into a 9-digit, non-conformant DATE on the last VEVENT of
year 9999. Confirmed at the source before fixing, and reproduced
against real `colitur emit --format ics --from 9999 --to 9999` output
(DTEND;VALUE=DATE:100000101) before touching any code.
RFC 5545 section 3.6.1: a VEVENT with a DATE-valued DTSTART and
neither DTEND nor DURATION has an implicit one-day duration, so
omitting DTEND for that one event is the standard's own correct
answer, not a workaround. dtend_of re-derives the successor's
year/month/day and re-validates them through Date.make -- the one
function that actually enforces the domain -- before trusting the
string; None means the caller omits the DTEND line entirely.
F2 (minor, same function): documented next_day's own Error branch as
dead-but-silent on shipped data (event's iso <> "" guard is the only
caller and always parses) -- behaviour unchanged, comment only.
Two new tests: the domain's last VEVENT (DTSTART 99991231) has no
DTEND line at all; every DTEND anywhere in a 9999 feed is exactly 8
digits (the general form of the bug, catches a regression anywhere
else in the domain too). Existing 2027/2028 DTEND-arithmetic
assertions untouched and still pass.
Mutation-proved: both new tests fail against the pre-fix code
(9-digit DTEND value caught verbatim), pass after.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Reuses resolved_year_report's existing two-liturgical-year indexing
rather than copying it: that walk owns the civil-vs-liturgical span
reasoning, and a second copy would drift. It is refactored to return the
days, with the printer layered on top, so day and readings behave
identically -- which cli.t proves byte-for-byte.
CSV emits one header for a whole multi-year run, not one per year.
A reversed range is a usage error rather than silently empty output.
Asserted in cli.t: two ics runs are byte-identical, because nothing in
the path reads a clock.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Not a template job: folding, escaping, exclusive DTEND and stable UIDs
are rules a logic-less template cannot enforce, and each fails silently
in a subscriber's client rather than loudly at generation.
DTEND is EXCLUSIVE for an all-day event (section 3.6.1). Wrong here
shows every event a day short, everywhere.
UIDs are YYYYMMDD-<rite>@colitur and stable across regenerations
(section 3.8.4.7). Wrong here duplicates the whole year in every
subscriber's phone, months later.
Every line is CRLF-terminated and folded at 75 octets (section 3.1).
No RRULE: a liturgical calendar is not a recurrence rule. Asserted, so
nobody optimises it later.
DTSTAMP is a parameter, not a clock read. RFC 5545 requires it and the
obvious implementation reads the wall clock -- which violates the
kernel's determinism rule and would make two feeds from identical data
differ byte-for-byte, defeating reproducible builds and any reviewable
diff on a published tree.
Corrected one test literal against real engine output: DTSTAMP is a
per-VEVENT property (section 3.8.7.2), not calendar-level, so the
default-value line count is 365 (every event), not 1.
Mutation-tested: a non-exclusive DTEND reddens the suite.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Element-per-field; attributes carry identity only and there is no mixed
content, so a consumer's XPath never has to distinguish the two.
Schema validation is an opt-in make check-schema via xmllint, not an
in-suite assertion: validating XSD needs an XML library and the
dependency list is frozen. It prints SKIPPED loudly when xmllint is
absent, because a silent skip reads as a pass. The suite asserts
well-formedness properties directly instead.
This corrects the design spec, which claimed in-test validation.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
view.ml's rank_label field was a byte-for-byte copy of the celebration's
name (names_value cel.Celebration.names), not a localized rank label at
all -- the kernel has no per-language rank names to draw one from, so
there was no honest value to put there. Nothing consumed it: no
template in the plan, no test, no other code referenced it.
Removed from both day_value and padding_cell so the two key sets stay
identical (23 keys each, verified). schema/day-v1.json already
described 23 keys and needed no change -- it now matches the emitted
output exactly.
schema/day-v1.json is a published contract: once a phone subscribes or
a site fetches this, removing a field is a breaking /v2/ change. The
time to remove a field that lies about its own contents is before
anyone can depend on it, not after.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Both consume the VIEW, not the kernel, so every emitter and every
template describe exactly the same fields -- there is one vocabulary,
not five.
CSV is RFC 4180: a field with a comma is quoted. That is live on real
data, not hypothetical -- 'St. Joseph, Spouse of the Bl. Virgin Mary'
would otherwise split into two columns.
JSON is hand-rolled because the dependency list is frozen and escaping
is the only subtlety. Control characters below 0x20 are \u-escaped per
RFC 8259 section 7.
There are no numbers in the view, deliberately: a consumer never has to
guess whether week is 2 or "2".
schema/day-v1.json pins the shape. Once a phone subscribes or a site
fetches this, it is a promise to strangers -- adding a field is minor,
renaming one means /v2/.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Shapes a civil year of resolved days into the value a template renders
against. This layer is why the engine can stay logic-less: a month grid
needs leading blank cells, week bucketing and an in-month test, and a
logic-less template can compute none of it.
Both weeks and days are offered at every level -- the booklet walks
days, the grid walks weeks -- so the two artefacts cannot drift.
Colours are six booleans, not hex: hex bakes a presentation policy into
the engine, and LaTeX, groff and HTML each want a different colour
expression. Asserted: exactly one of the six is true on every day of a
whole year, so a template keying off them can never get none or two.
Padding cells carry every field a real day carries, empty, so a template
never hits a missing key mid-grid.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Every interpolated value is escaped for the template's flavour; the
template's own literal text never is, because that is the author's
markup. There is no raw form, so a template cannot opt out.
Scope is a stack with outward fallback, so a grid template can reach the
year number from inside a week without the view duplicating it into
every cell.
A missing key renders empty -- the one deliberate silence, so a template
survives a rite that does not set every optional field.
Mutation-tested: dropping the Escape.apply call reddens the
data-cannot-escape-flavour case.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
F1: test_no_raw_or_partial_form's first assertion only excluded one
literal shape (Ok [Var ["{name"]]), so it could not actually catch a
future raw/unescaped constructor under a different name. Replace it
with an assertion of the real parse result for {{{name}}}
(Ok [Var ["{name"]; Text "}"]), documented behaviour rather than a
guarantee this test cannot check -- the real guarantee is structural:
node has exactly four constructors and none of them is raw.
F2: {{.}}, {{#}}, {{^}} and {{/}} used to parse to a Var/Section/
Inverted with an empty path, reachable but never designed. This engine
has no "current context" for a bare dot to mean, so a bare-dot or
empty-sigil path is now a parse error at lex time, covering all four
sigil forms via one path helper. The existing "empty tag {{}}" branch
is unchanged and still reachable (a fully empty body is a distinct
case from a sigil with an empty path).
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Placeholders, sections, inverted sections, comments. Nothing else: no
partials, no lambdas, no expression evaluation, no raw form. A template
is data, never a program, which is what keeps an untrusted template safe.
Errors rather than silence on a malformed template: an unterminated tag,
an unclosed section, a mismatched close and a partial all return Error.
Swallowing '{{name' as text is how a typo becomes invisible missing
output in a printed booklet.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
fold_ics's UTF-8 backoff loop could back `cut` all the way down to
`pos` on 74+ consecutive continuation bytes (0x80-0xBF), producing a
zero-length chunk and recursing on the identical position forever --
not producible by valid UTF-8, whose longest continuation run is 3,
but the kernel's own totality requirement covers arbitrary octet
strings, not only valid ones. When backoff finds no boundary inside
the window, cut hard at the limit instead, so forward progress is
unconditional.
test_fold_never_splits_utf8 previously asserted only that unfolding
reproduced the original bytes, a property folding preserves at any
cut position and therefore blind to a boundary violation. It now also
asserts the named property directly: no continuation chunk may start
with a UTF-8 continuation byte. A new regression test feeds fold_ics
100 consecutive continuation bytes and asserts it terminates with
every line at or under 75 octets.
|
