Urlwatch

Added

- New filter: `pretty-xml` to indent/pretty-print XML documents
- New filter: `jq` to parse, transform, and extract JSON data
- New reporter: `prowl` (by nitz)

Fixed

- Proper multi-line highlighting for wdiff (PR615, by kongomongo)
- Fix command-line generation for html2text (PR619, by Eloy Paris)

Added

- Added 'wait_until' option to browser jobs to configure how long
the headless browser will wait for pages to load.
- Jobs now have an optional `treat_new_as_changed` (default `false`)
key that can be set, and will treat newly-found pages as changed,
and display a diff from the empty string (useful for `diff_tool`
or `diff_filter` with side effects)
- New reporters: `discord`, `mattermost`
- New key `user_visible_url` for URL jobs that can be used to show
a different URL in reports (useful if the watched URL is a REST API
endpoint, but the report should link to the corresponding web page)
- The Markdown reporter now supports limiting the report length via the
`max_length` parameter of the `submit` method. The length limiting logic is
smart in the sense that it will try trimming the details first, followed by
omitting them completely, followed by omitting the summary. If a part of the
report is omitted, a note about this is added to the report. (PR572, by
Denis Kasak)

Changed

- Diff output is now generated more uniformly, independent of whether
the input data has a trailing newline or not; if this behavior is not
intended, use an external `diff_tool` (PR550, by Adam Goldsmith)
- The `--test-diff-filter` output now properly reports timestamps from
the history entry instead of the current date and time (Fixes 573)
- Unique GUIDs for jobs are now enforced at load time, append "1",
"2", ... to the URLs to make them unique if you have multiple
different jobs that share the same request URL (Fixes 586)
- When a config, urls file or hooks file does not exist and should be
edited or inited, its parent folders will be created (previously
only the urlwatch configuration folder was created; Fixes 594)
- Auto-matched filters now always get `None` supplied as subfilter;
any custom filters must accept a `subfilter` parameter after the
existing `data` parameter
- Drop support for Python 3.5

Fixed

- Make imports thread-safe: This might increase startup times a bit,
as dependencies are imported on bootup instead of when first used.
Importing in Python is not (yet) thread-safe, so we cannot import
new modules from the worker threads reliably (Fixes 559, 601)

- The Matrix reporter was improved in several ways (PR572, by Denis Kasak):

- The maximum length of the report was increase from 4096 to 16384.
- The report length limiting is now implemented via the new length limiting
functionality of the Markdown reporter. Previously, the report was simply
trimmed at the end which could break the diff blocks and make them render
incorrectly.
- The diff code blocks are now tagged as diffs which will allow the diffs to
be syntax highlighted as such. This doesn't yet work in Element, pending on
the resolution of trentm/python-markdown2370.

Added

- Added `--test-reporter REPORTER` command-line option to send an
example report using any configured notification service
- `JobBase` now has `main_thread_enter()` and `main_thread_exit()`
functions that can be overridden by subclasses to run code in
the main thread before and after processing of a job
(based on an initial implementation by Chenfeng Bao)

Removed

- The `--test-slack` command line option has been removed, you can
test your Slack reporter configuration using `--test-reporter slack`

Changed

- The `browser` job now uses Pyppeteer instead of Requests-HTML for
rendering pages while executing JavaScript; this makes JavaScript
execution work properly (based on code by Chenfeng Bao)

Fixed

- Applying legacy `hooks.py` filters (broken since 2.19; reported by Maxime Werlen)

Added

- A job can now have a `diff_filter` set, which works the same way as the normal
`filter` (and has the same filters available), but applies to the `diff` output
instead of the page content (can be tested with `--test-diff-filter`, needs 2
or more historic snapshots in the cache)
- Documentation now has a section on the configuration settings (`--edit-config`)
- New filter: ``ocr`` to convert text in images to plaintext (using Tesseract OCR)
- New reporters:
- ``ifttt`` to send an event to If This Then That (ifttt.com) (512, by Florian Gaultier)
- ``xmpp`` to send a message using the XMPP (Jabber) protocol (533, by Thorben Günther)

Changed

- The `urlwatch` script (Git only) now works when run from different paths
- Chunking of strings (e.g. for Slack and Telegram) now adds numbering (e.g.
` (1/2)`) to the messages (only if a message is split into multiple parts)
- Unit tests have been migrated from `nose` to `pytest`
and moved from `test/` to `lib/urlwatch/tests/`
- The ``css`` and ``xpath`` filters now accept ``skip`` and ``maxitems`` as subfilter
- The ``shellpipe`` filter now inherits all environment variables (e.g. ``$PATH``)
of the ``urlwatch`` process

Fixed

- The ``html2text`` method ``lynx`` now treats any subfilters with a non-``null``
value as command-line argument ``-key value`` (previously only the value ``true``
was treated like this, and any other values were silently dropped)

Added

- Documentation is now available at [urlwatch.readthedocs.io](https://urlwatch.readthedocs.io)
and shipped in the source tarball under `docs/`; filter examples in the docs are unit-tested
- New filters:
- `reverse`: Reverse input items (default: line-based) with optional `separator`
- `pdf2text`: Convert PDF files to plaintext (must be first filter in chain)
- `shellpipe`: Filter text with arbitrary command-line utilities / shell scripts
- `FilterBase` API improvements for specifying subfilters:
- Add `__supported_subfilters__` for sub filter checking and `--features` output
- Add `__default_subfilter__` to map value-only parameters to dict parameters,
for example the `grep` filter now has a default subfilter named `re`
- Support for using Redis as a cache backend via `--cache=redis://localhost:6379/`

Fixed

- Declare updated Python 3.5 dependency in `setup.py` (already a requirement since urlwatch 2.18)

Changed

- Filter improvements:
- `sort`: Add `reverse` option to reverse the sorting order
- `sort`: Add `separator` option to specify item separator (default is still line-based)
- `beautify`: The `jsbeautifier` (for `<script>` tags) and `cssbeautifier` (for `<style>` tags)
module dependencies are now optional - if they are not installed, beautify only works on the HTML
- Most filters that only had unnamed subfilters (e.g. `grep`) now have a named default subfilter
- Reporter improvements:
- ``pushover``: The message ``priority`` can now be configured
- Travis CI: Set `pycodestyle` version to 2.6.0 to avoid CI breakage when new style checks are added
- Diff results are now runtime cached on a per-job basis, which shouldn't affect behavior, but
could be observed by an external `diff_tool` running at most once per job instead of multiple times
- Jobs with a custom `diff_tool` or a `shellpipe` filter are now ignored if `jobs.yaml` has the
world-writable bit (`o+w`) set or is not owned by the current user (does not apply to Windows);
previously only `shell` jobs were ignored if the permissions/owners were wrong

Deprecated

- String-based filter definitions (e.g. `html2text,grep:Current.*version,strip`) are now
deprecated, it is recommended to use YAML-based dictionary-style listing of filters,
which is more flexible, easier to read and write and more structured

Added
- New filter: `re.sub` that can replace/remove strings using regular expressions
- Support `ignore_timeout_errors` and `ignore_too_many_redirects` for URL jobs (423, by Josh aka Zevlag)
- HTML reporter: Add `viewport` meta tag for improved viewing on mobile devices (432, by Mike Borsetti)
- Optional support for insecure SMTP password storage in the config; use with caution (431)
- Add `matrix` reporter
- New filter: `beautify` that can beautify HTML, JavaScript and CSS

Fixed
- Fix `--test-filter` when the specified job is not found
- Fix another `YAMLLoadWarning` in unit tests (382, by Louis Sautier)
- Documentation updates and typo fixes (by Nate Eagleson)
- Pushover: Fix default device config (Fixes 409 and 372, documented by Richard Goodwin)

Changed
- Nicer formatting of `--features` for jobs with no docstring or many keys
- The XPath and CSS filters now support XML namespaces (404, by Chenfeng Bao)
- Drop support for Python 3.3 and Python 3.4 (new minimum requirement is Python 3.5)
- Use `html.escape` instead of `cgi.escape` (which was removed in Python 3.8; 424, by Chenfeng Bao)
- Allow non-ASCII characters in `format-json` output filter (433, by Mike Borsetti)
- The `keyring` config option for `email` was changed to `auth`; if you have problems
with authentication where none is needed, set `report/email/smtp/auth` to `false`