Regression from #62: verify_ssl true used to resolve to the system CA
bundle path, so corporate CAs installed in the OS worked; after the
unification true became requests' default (certifi), breaking setups
with a corporate CA in the system store.
Now verify_ssl true injects truststore, so requests verifies against
the OS trust store on any platform. verify_ssl false / custom CA path
behavior is unchanged. Tests mock truststore via an autouse fixture to
keep the pytest process free of global ssl mutation.
Refs #62
- README.md: 324 -> 106 lines, clickable TOC, quick start, formats
table, full check suite + CI; flags/env/YAML reference moved to docs
- docs/USER_GUIDE.md: new user guide (install, periods, users, export,
email, monthly --commit cycle, 21-flag reference, troubleshooting)
- docs/CONFIG.md: clickable TOC + accuracy fixes verified against code
(dedup logic was described backwards, missing report section in
--init-config, conditional STARTTLS, full --config-path role,
datetime --date ranges, --config override nuance, comment loss on
--commit, default_to without default_from is ignored)
- client.py: dedup docstring/comment now match actual behavior
- .gitea/workflows/checks.yaml: runs on push to main and every PR,
matrix Python 3.10-3.13 (matches requires-python >=3.10), steps:
isort, black, ruff check, ruff format, mypy, pytest
- dev extras gain black, isort (needed by pip install -e .[dev] in CI)
and setuptools>=61.0 (test_formatters_found_by_setuptools imports
find_packages at runtime; Python 3.12+ venvs no longer bundle it)
- README gains pipeline status badge
Closes#53
isort --check-only failed on 6 files because the project had no isort
configuration and its default style conflicts with black/ruff-format.
With profile = "black" all formatters agree; no source changes needed.
Регрессия от 1df1194 (#64): load_dotenv переехал с module-level в
Config.load_yaml(), но --init-config выходит до этого вызова, поэтому
_run_init_config больше не видел значения из .env — пользователь с
настройками только в .env получал пустой YAML, хотя docs/CONFIG.md
обещает чтение текущих значений из .env.
В начале _run_init_config добавлен load_dotenv(find_dotenv(usecwd=True),
override=False): поиск .env идёт от текущей директории (plain-вызов без
usecwd стартует от директории cli.py и .env пользователя не находит),
переменные окружения не перебиваются. Импорт модуля по-прежнему без
side effects — вызов внутри функции.
Регрессионный тест: test_init_config_reads_dotenv_file (реальный .env
в tmp_path + chdir, без мока os.environ на целевые переменные).
Refs #64
cli.py: drop redundant 'if issue_hours is None' branch before
'if not issue_hours'. The branch is not formally dead (client
returns None when no time entries found, client.py), but it is a
semantic duplicate: both branches print the same message and
return 0, and None is falsy, so a single 'not issue_hours' check
covers None, empty list and keeps behavior identical.
test_cli_returns_zero_on_no_entries stays green unchanged.
xlsx.py: drop unreachable 'if ws is None' fallback. Workbook()
(write_only=False) always creates one worksheet in __init__, so
wb.active is never None; set the title directly.
Closes#63
Без --date, env и YAML отчёт строится за текущий месяц: начало периода —
1-е число текущего месяца, конец — сегодня (date.today()), согласовано с
fallback default_to→today из #50. Приоритеты источников
(CLI > env > .env > YAML > дефолт) и precision=datetime не затронуты.
Сам fallback уже существовал в коде (хардкод устаревшего периода был
убран ранее); коммит фиксирует поведение детерминированными тестами с
моком date.today (RED проверен регрессионным зондом: при возврате
хардкода 2025-12-19--2026-01-31 тесты падают) и обновляет README и
docs/CONFIG.md: дефолт явно описан как «текущий месяц».
Closes#56
Импорт модулей больше не мутирует процесс:
- config.py: module-level load_dotenv(override=False) мутировал os.environ
при любом импорте. Вызов перенесён в Config.load_yaml() — точку явной
загрузки конфигурации, семантика override=False и путь по умолчанию
сохранены. --config PATH (load_dotenv(path, override=True)) не затронут.
- mailer.py: глобальная регистрация email.charset.add_charset('utf-8', ...)
меняла кодировку для всего процесса. Заменена на точечное применение
8bit UTF-8 к телу письма в _utf8_text_part() — payload в UTF-8
(surrogateescape) + Content-Transfer-Encoding: 8bit на конкретной части.
Проверено: сериализация в байты (BytesGenerator, как в send_message)
побайтово совпадает со старым поведением.
Тест test_env_var_takes_priority_over_dotenv адаптирован к новой точке
вызова load_dotenv; test_body_substitution переведён с as_string() на
декодированный payload (str-сериализация не-ASCII части без глобальной
мутации реестра теперь даёт base64; wire-формат 8bit сохранён и
зафиксирован test_wire_format_is_8bit_utf8).
Closes#64
YAML verify_ssl: true раньше подставлял захардкоженный путь
/etc/ssl/certs/ca-certificates.crt, а REDMINE_VERIFY=true — bool True.
Путь отсутствует на части дистрибутивов, семантика источников различалась.
Теперь едино для YAML и env:
- true (bool/строка) → True: стандартная проверка TLS средствами requests;
- false (bool/строка) → False (+ сохраняется warning из #57);
- иная строка → путь к CA-bundle как есть.
- дефолт (значение не задано) → True вместо захардкоженного пути.
Существующие тесты на путь-от-true переписаны под новую семантику
(изменение поведения): test_verify_ssl_true_returns_default_ca_path →
test_verify_ssl_true_returns_true.
Closes#62
Redmine/requests exceptions may embed the request URL containing the
API key (?key=SECRET) or the X-Redmine-API-Key header value. The CLI
printed the raw exception text to stderr, leaking the key into
terminals and logs.
Add _sanitize_error_text() which masks 'key=<value>' query params and
the X-Redmine-API-Key header as 'key=***' / 'X-Redmine-API-Key: ***',
and apply it to every exception-derived message printed to stderr
(fetch errors in main() and SMTP errors in _save_and_maybe_send()).
Full traceback with original exception details is now available under
both --verbose and --debug (previously only --debug, only for
RedmineAPIError). Exit codes and error-handling structure unchanged.
Closes#55
REDMINE_VERIFY=false / verify_ssl: false silently disabled TLS
certificate verification, exposing the connection to MITM attacks.
Config.validate() now prints a warning to stderr when verification
is disabled (exactly False; True or a CA-bundle path stays quiet).
validate() runs exactly once at CLI startup, so the warning is
emitted once per run and is visible in every scenario.
Closes#57
Config.validate() now rejects REDMINE_URL values whose scheme is not
HTTPS (http://, missing scheme, or any other scheme). The API key is
sent in request headers, so a plain-HTTP endpoint would expose it.
Scheme comparison is case-insensitive per RFC 3986.
Closes#54
Issues not returned by issue.filter (no access / deleted) and failures
of the time entry activities lookup were silently swallowed: the report
was built from partial data without telling the user.
Now both cases emit a warning to stderr (project ⚠️ style, #28):
- skipped issue IDs and total lost hours after issue matching;
- activities lookup failure (still returns {} and falls back to
activity names from time entries).
These are warnings, not errors: the report is still built from
available data and the exit code is unchanged.
Closes#61
Redmine user.filter(login=...) performs an inexact substring search, but
_resolve_user_id took users[0].id unconditionally, so a report could
silently be built for the wrong user (e.g. 'ivanov' matching 'ivanov2').
Now only exact (case-sensitive) login matches are considered: exactly one
match resolves to its id, multiple matches raise an ambiguity error,
no match falls through to the name lookup and then to a 'not found'
error — symmetric with the --user-name resolution.
Closes#60
Dedup with precision=datetime crashed with TypeError: redminelib returns
naive created_on/updated_on while the cutoff from _compute_dedup_cutoff
is aware UTC. Normalize at a single point: _parse_datetime now always
returns an aware datetime (naive treated as UTC), matching its docstring.
The dedup cutoff is normalized the same way, and any residual comparison
TypeError is wrapped in RedmineAPIError instead of leaking raw.
Also fix --commit saving last_used.to as naive local time; it now stores
aware UTC (datetime.now(timezone.utc)) so the next run computes a correct
aware cutoff.
Closes#58
With period.dynamic: true and period.precision: datetime, running
without --date crashed: compute_next_period() returns a datetime range
(YYYY-MM-DDTHH:MM:SS), which parse_date_range() rejected with a
fullmatch against YYYY-MM-DD only.
parse_date_range now accepts YYYY-MM-DDTHH:MM:SS--YYYY-MM-DDTHH:MM:SS
in addition to plain date ranges, returning datetime strings unchanged
(pass-through, matching compute_next_period output). Mixed-precision
ranges, invalid times and reversed ranges are still rejected with the
existing error messages; date-range behavior is unchanged.
Closes#59
The trap mocked issue.filter to return only issue 2, so result == [2]
would pass even with a wrong fix that suppresses TypeError but keeps all
entries. Now the API mock returns both issues; only correct filtering
of the pre-cutoff time entry keeps the result at [2].
Refs #67, #58
- tests/test_cli.py: fix 26 fetch_issues_with_spent_time mocks to return
3-element tuples (issue, hours, activities) matching the real signature
- pyproject.toml: add [tool.pytest.ini_options] with testpaths and
pythonpath so bare pytest works
- tests/test_client.py: add pagination test (>100 time entries arriving
in pages, hours aggregated across all pages)
- tests/test_formatters.py: add structural tests — XLSX full header row
and grand total row via openpyxl, HTML thead with all columns and
tbody row count
- add strict xfail trap-tests for known bugs: #58 (naive created_on vs
aware dedup cutoff TypeError) and #59 (parse_date_range rejects
datetime range from dynamic+datetime period)
Closes#67
redmine_reporter/client.py imports requests and urllib3 directly but
they were only available transitively via python-redmine. Retry with
allowed_methods requires urllib3 >= 1.26.
Closes#66
pyproject.toml declared requires-python >= 3.9, but the codebase uses
annotations that are incompatible with Python 3.9: builtin generic
tuple[str, str] in parse_date_range (cli.py) and PEP 604 unions
str | None (config.py, yaml_config.py), the latter requiring 3.10+
at runtime.
Bump requires-python to >=3.10 and drop the Python 3.9 classifier.
Closes#51
Add email.html config flag (default false). When enabled, --send
includes an HTML version of the report body generated via HTMLFormatter
alongside the plain-text part in a multipart/alternative message.
- EmailConfig gains html: bool field
- mailer.build_message/send_report accept rows for HTML generation
- CLI passes rows to send_report
- --init-config generates email.html: false
- README.md and docs/CONFIG.md updated
Bump version to 1.10.0.
Closes#48
Add YAML section `report.no_time` that controls `--no-time` behavior
in automatic modes (`--commit`, `--send`). CLI flag `--no-time`
always wins. Manual `--output` ignores the YAML setting.
- Config.get_report_no_time() reads the YAML value (defaults to false)
- cli._resolve_no_time() encapsulates CLI > YAML priority
- --init-config now generates the report section
- docs updated in README.md and docs/CONFIG.md
Closes#49
- README: add --send flag to features, usage examples, full flag list;
replace black/isort with ruff in dev section; add email config
template variables docs
- CONFIG: new email section with all fields documented, --send usage
examples, error handling and flag compatibility table
- pyproject.toml: remove unused black and isort from dev dependencies
and their tool configs (project uses ruff for both lint and format)
- cli.py: wrap save_period_to_config in try/except to avoid
unhandled traceback on disk full / permission denial (I2)
- test_cli.py: remove duplicate _MockIssue class definition (I3)
Closes#45
- New redmine_reporter/mailer.py: SMTP email sending with
{author}/{period} template substitution, MIME attachment
with correct content-type per file extension
- Config.get_email_config(): returns EmailConfig from YAML
or None when not configured
- CLI --send flag: sends report after generation, works with
--output, --commit, or standalone (saves to template path)
- 31 new tests (22 mailer + 6 CLI + 3 config)
- 249/249 tests passing, ruff clean, mypy clean
- --commit: после генерации отчёта сохраняет период в
period.last_used.from/to YAML-конфига
- При period.dynamic=true следующий запуск вычисляет период от last_used:
полный месяц → следующий месяц, произвольный диапазон → та же длина
- При period.dynamic=false перезаписывает default_from/default_to
- При period.precision=datetime сохраняет timestamps с точностью до секунд
- --commit без --output сохраняет отчёт по шаблону из конфига
- compute_next_period() в config.py
- save_period_to_config() в yaml_config.py
- 23 новых теста (7 CLI, 6 config, 6 yaml, 4 date_range)
Closes#44
#43 — Имя файла и пути по умолчанию:
- resolve_output_path() в yaml_config.py: резолвит --output с учётом
output.dir, filename_template, default_format из YAML-конфига
- Bare format (xlsx/odt/...) → путь по шаблону
- Без расширения → автодописывается default_format (.xlsx)
- Config.get_output_dir/filename/default_format()
#47 — datetime precision и дедупликация:
- period.last_used.from/to в YAML-конфиге и AppConfig
- period.precision: date|datetime
- _compute_dedup_cutoff() в cli.py: при precision=datetime вычисляет
cutoff из period.last_used.to
- _parse_datetime() + AND-логика дедупликации в client.py:
запись исключается если created_on И updated_on < cutoff
- Пропуск issues без часов после дедупликации
Closes#43Closes#47
Previously boolean True was passed directly to python-redmine
which used system CA bundle instead of the project's default CA file.
Now 'true' in YAML maps to DEFAULT_REDMINE_VERIFY path.
The template contained an empty <text:p/> that was rendered as
a blank line before the report header. After loading the template,
strip all text:p children so that ODT output no longer depends on
template editing artifacts.
Closes#42.
- Load time entry activity names from Redmine enumeration.
- Aggregate hours per issue and per activity in fetch_issues_with_spent_time.
- Add --by-activity CLI flag and propagate it through report builder,
summary, and all formatters (console, CSV, HTML, JSON, ODT).
- Keep backward-compatible 2-tuple input in build_grouped_report.
- Bump version to 1.8.0.
Closes#41
- Add user_id parameter to fetch_issues_with_spent_time().
- Support numeric ID, login, or full name resolution.
- Reject ambiguous names and unknown users with clear messages.
- Add CLI flags: --user-id, --user-login, --user-name.
- Only allow one user flag at a time.
- Add ResourceNotFoundError handling.
- Update README with usage examples.
- Add tests for user resolution and CLI flags.
Closes#40
