Some Unicode uppercase letters (e.g., mathematical bold 𝐀) have no
distinct lowercase form, so keyword.lower() == keyword. Filter these
out since the test requires a genuine case difference.
- Update hypothesis to latest version (>=6.151.9)
- Remove pythonpath pytest config (was only needed for tests.strategies)
- Consolidate classic_theme.py into single file with all design models
- Move Hypothesis strategies from strategies.py into their test files
- Add noqa: ARG001 to unused yaml_field_override CLI parameter
- Fix lint and type errors across the codebase
The regex in remove_not_provided_placeholders used a bare alternation
without word boundaries, so removing placeholder "AA" would also
destroy "AAA" in the same template. Fix: add \b word boundaries and
sort longest-first.
Also add Hypothesis tests for:
- remove_not_provided_placeholders: provided keys survive removal
- remove_connectors_of_missing_placeholders: connectors removed/preserved
- validate_arbitrary_date: pass-through for valid dates and custom text
- Mastodon URL: domain and username appear in generated URL
The \b word boundary anchor requires a word character (\w) on one side
and a non-word character on the other. Keywords ending with ":" or
starting with "+" have no word boundary at those positions, so \b
fails to match.
Fix: only add \b on the side of a keyword where the character is a
word character. For "Tech stack:", \b is added before "T" but not
after ":". For "C++", \b is added before "C" but not after "+".
Added a Hypothesis property test that catches this entire class of
bugs: any keyword surrounded by spaces in the input text must appear
bolded in the output.
Removed tests that were running 100 random inputs through string
operations that can never fail regardless of input:
- process_highlights: fuzzing "- " + string prepend
- process_authors: fuzzing ", ".join()
- PublicationEntry DOI URL: fuzzing f"https://doi.org/{doi}"
These tests add execution time without any chance of finding bugs.
Test that random valid DOI strings always produce the correct
https://doi.org/ URL. Removed the useless test_doi_clears_url that
was just re-verifying a static validator with random irrelevant input.
- date_only_clears_start_and_end: providing date nullifies start/end
- start_only_implies_present: start without end defaults to "present"
- end_only_becomes_date: end without start treated as single date
Merge property-based tests into existing test classes per the project
testing rules: multiple tests for the same function belong in a
TestFunctionName class.
- TestEscapeTypstCharactersProperties merged into TestEscapeTypstCharacters
- TestMarkdownToTypstProperties merged into TestMarkdownToTypst
- TestBuildDatePlaceholdersProperties merged into TestBuildDatePlaceholders
- TestComputeTimeSpanStringProperties merged into TestComputeTimeSpanString
- TestGetDateObjectProperties renamed to TestGetDateObject
- TestApplyOverridesProperties merged into TestApplyOverridesToDictionary
- TestUpdateValueByLocationProperties merged into TestUpdateValueByLocation
- TestResolveOutputFolderPlaceholderProperties merged into TestResolveOutputFolderPlaceholder
- TestBuildNameVariantsProperties renamed to TestBuildNameVariants
- TestTypstDimensionProperties renamed to TestTypstDimension
- TestSocialNetworkUsernameProperties merged into TestSocialNetwork
Move all 55 property-based tests from tests/test_hypothesis.py into
the existing test files for each module they test:
- test_string_processor.py: keyword bolding, placeholder, URL, pattern
- test_markdown_parser.py: Typst escaping, markdown-to-typst
- test_date.py: date parsing, placeholders, time spans
- test_override_dictionary.py: immutability, path traversal
- test_path_resolver.py: name variants, OUTPUT_FOLDER resolution
- test_classic_theme.py: Typst dimension validation
- test_social_network.py: username format validation
Reusable Hypothesis strategies live in tests/strategies.py. Added
pythonpath=["."] to pyproject.toml so tests can import the strategies
module.
1. clean_url: use rstrip("/") instead of single-slash strip. URLs
with multiple trailing slashes (e.g., "example.com//") now clean
fully. Also restores idempotency (clean(clean(x)) == clean(x)).
2. get_date_object: use Date(year, 1, 1) instead of
Date.fromisoformat(f"{date}-01-01"). The fromisoformat call
crashed on years < 1000 because it requires 4-digit year strings.
3. build_date_placeholders: use f"{year % 100:02d}" for
YEAR_IN_TWO_DIGITS instead of str(year)[-2:]. The slice produced
1-char output for single-digit years (e.g., year 9 gave "9"
instead of "09").
Test ranges widened back to full datetime.date range now that the
source code handles all values correctly.
Extract three modules from classic_theme.py (845 lines) to improve
navigability:
- templates.py: entry template option models and Templates class
- typography.py: Typography, FontFamily config, FontSize, SmallCaps, Bold
- header.py: Header, Connections, Links, PhoneNumberFormatType
classic_theme.py retains Page, Colors, Sections, Entries, and
ClassicTheme. All imports use absolute paths so the custom theme
generator (which copies classic_theme.py as text) continues to work.
Template entry option classes renamed from e.g. OneLineEntry to
OneLineEntryTemplate to avoid name collision with CV entry models.
- Revert entry_with_date.py to let ValueError propagate naturally
(Pydantic produces better messages like "The month must be between
1 and 12." than our generic wrapper)
- Update create_theme tests to expect typer.Exit (now has error handler)
- Update pdf_png test to use real temp file instead of patching read_text
- Fix watcher test mock to handle double join() call pattern
- Regenerate JSON schema after union type order change
- Use Path.open() instead of open() in pdf_png.py (PTH123)
- Chain KeyError with 'from e' in pdf_png.py (B904)
- Fix comment placement in section.py to avoid false unused-ignore
- Update watcher test to mock observer.join() instead of time.sleep()
Add a cached_property that maps each language to its primary country's
flag emoji using regional indicator symbols, with tests and updated
developer guide.
The rich progress panel outputs ✓ (U+2713), which Windows cp1252
console encoding cannot encode. Set PYTHONIOENCODING=utf-8 in the
subprocess environment.
Bundle the typst-fontawesome package as a git submodule so that Font
Awesome icons are available during Typst compilation without network
access. Refactor get_package_path to install multiple bundled packages
into the temporary cache. Add an offline wheel installation test that
renders a PDF with a poisoned HTTP proxy to verify no network calls.
Bump black (26.3.1), ruff (0.15.7), ty (0.0.24), prek (0.3.6), typer
(0.24.1), and codespell (v2.4.2). Add ty:ignore comments and type
annotations to satisfy stricter checks in ty 0.0.24. Make skill zip
generation reproducible with a fixed timestamp.
When using CLI overrides like `--design.theme "moderncv"` with a separate
design file, `update_value_by_location` crashed with `KeyError: 'design'`
because intermediate dictionary keys were not auto-created during traversal.
Now missing intermediate keys are initialized as empty dicts, allowing
overrides to work regardless of whether the target section exists in the
main YAML.
process_summary() unconditionally wrapped summary text in Markdown admonition
syntax (!!! summary), which only works as a block-level construct on its own
line. When users placed SUMMARY inline in a template (e.g., "**COMPANY**,
SUMMARY"), the admonition syntax broke rendering.
Now render_entry_templates() checks if SUMMARY appears on its own line in
the template. If standalone, admonition wrapping is applied for the special
#summary[] Typst formatting. If inline, the raw summary text is used directly.
When users specified design or locale overlay file paths in their YAML via
settings.render_command.design or settings.render_command.locale, these were
parsed but never actually loaded as overlays. Only CLI flags (--design,
--locale) triggered overlay loading.
Now cli_command_render resolves these paths from the YAML (via the existing
collect_input_file_paths helper) before building arguments, so overlay files
referenced in settings are loaded the same way as CLI-provided ones.
The FontFamily JSON schema used a strict enum that rejected custom and system
fonts like Georgia. Changed from SkipJsonSchema[str] to plain str in the union
type, so the generated schema uses anyOf with both the known font enum (for
editor autocomplete) and a generic string type (for arbitrary fonts).
On NixOS and other immutable distributions, package files in the store are
read-only. shutil.copytree preserves source permissions, so copied template
files remain read-only. This causes PermissionError when create-theme tries
to write __init__.py into the copied directory.
Added make_tree_writable() that adds user-write permission to all files and
directories after copying. Called automatically by copy_templates().
When multi-line template fields (like main_column) contained markdown emphasis
markers on adjacent lines, the Python markdown library treated single-newline-
separated lines as one paragraph, causing markers to interact across lines and
produce garbled Typst output with mismatched brackets.
The fix processes each line of input independently in markdown_to_typst() while
preserving multi-line admonition blocks. This prevents emphasis markers on one
line from interfering with markers on adjacent lines.
Ink: elegant academic theme with EB Garamond and purple palette.
Opal: modern minimalist with Lato and teal accents.
Ember: editorial theme with mixed Gentium Book Plus/Ubuntu fonts,
crimson accents, diamond bullets, and justified-with-no-hyphenation.
Updates README theme grid from 2x3 to 3x3 to showcase all 9 themes.
The skill file lived in the main repo, which meant `npx skills add`
had to clone the entire RenderCV codebase and exposed internal dev
skills. A dedicated lightweight repo solves both problems.
- Create rendercv/rendercv-skill repo as a read-only distribution channel
- Add it as a submodule at .claude/skills/rendercv-skill/
- Update generation script and tests to write to the submodule path
- Add submodules: true to test workflow checkout steps
- Update docs, README, and install commands to use rendercv/rendercv-skill
- Add --recursive clone instruction to developer guide
- Delete the old skills/ directory at repo root
- Add SKILL.md for AI agents: YAML schema reference, entry type field
tables, design samples per theme, CLI reference, and important patterns
(YAML quoting, phone validation, bullet characters)
- Add Jinja2 template and generate.py that auto-generates SKILL.md and
docs/llms.txt from live Pydantic models and sample generators
- Add promptfoo eval suite (15 tests across 4 files): cv_generation,
pdf_parsing, design, and cli_workflow
- Add deterministic grader that validates LLM output through RenderCV's
own pydantic pipeline (not jsonschema)
Move `rendercv-typst/` into `src/rendercv/renderer/rendercv_typst/` so that
`lib.typ` and `typst.toml` ship with every pip install. This eliminates the
runtime dependency on Typst Universe, allowing releases without waiting for
external PR reviews. The wheel excludes non-essential files (examples, template,
README, etc.) via `[tool.uv.build-backend]` wheel-exclude.
Add four new section title styles: `centered_without_line`,
`centered_with_partial_line` (baseline), `centered_with_centered_partial_line`
(middle-aligned), and `centered_with_full_line`. These allow section headings
to be centered with various line decorations, complementing the existing
left-aligned options.
When a placeholder like DEGREE was missing from an education entry, the
locale phrase "DEGREE in AREA" left the connector word "in" behind,
producing output like "Princeton University, in Computer Science". This
affected all locales with connector words (en, em, di, at, etc.).
The fix adds a general-purpose `remove_connectors_of_missing_placeholders`
step inside `remove_not_provided_placeholders` that strips bare connector
words from separators between placeholders when either side is missing.
This works for any template, not just locale phrases — for example,
"JOB_TITLE at COMPANY" now correctly drops "at" when COMPANY is absent.
