* do better with loop cleanup
* changelog
* remove redundant line
* Do this a bit better than the initial pass
* Improve windows support
Make some other things coroutines to work with improved design
* Wish we'd have done this right from the start...
* Update deps surrounding this
- see bpo-23057
- neccessary for windows users
- nice for consistent support channel info / feature availability
* dep issue
* Fix tests
* duplication plugin py version
* actually handle this
* Reconfigure some checks with codeclimate, disable pylint for now
* style
* Is my exasperation showing yet?
* handle some stupid stuff
* meh
* dep changelog
* create
* add the code
* change de logs
* lets remember linux exists
* fix formtting and change wording
* change the wording a bit more... after running `sphinx-build` it def looks like i used bold a lot
god that was a long commit mesage
* review changes
* draper's review (see description)
i have put or worse partially to scare users and partially as there are other possible effects of a downgrade
* review
* clarity
* docs: change links to d.py docs to use stable version instead of v1.0.1
* chore(changelog): add towncrier entry
* docs: add |DPY_VERSION| substitution and :dpy_docs: role
* chore(changelog): update towncrier entries to reflect new changes
* commit
* add launcher depricate to linix/mac
* sorry linux i still ignored you
capitalise some linix stuff i couldnt be bothered to do before
* remove mentions of launcher & update changelog enrty
* [Docs] Update linux install docs, redo venv docs
Some of our pre-req installation docs needed updating on Windows - this adds new sections for Fedora Linux and Debian/Raspbian Buster, and also removes some unnecessary pre-requirements from other distributions. These have all been tested on fresh VPSes, installing Red both in venvs and with --user, and they all seem to work.
Also, apparently the venv docs were too scary before. These changes try to make it clear that it's easier to use than users may think.
This also includes a little note to stop users from accidentally installing Python with pyenv after installing pre-requirements on Ubuntu.
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Add changelog entries
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Update officially supported platforms in README.md
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Combine sections and add openSUSE
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Include example of using `--user` flag
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Use `py -3.7` on Windows outside of venv
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Reorganise changelog entries
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* PostgreSQL driver and general drivers cleanup
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Make tests pass
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Add black --target-version flag in make.bat
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Rewrite postgres driver
Most of the logic is now in PL/pgSQL.
This completely avoids the use of Python f-strings to format identifiers into queries. Although an SQL-injection attack would have been impossible anyway (only the owner would have ever had the ability to do that), using PostgreSQL's format() is more reliable for unusual identifiers. Performance-wise, I'm not sure whether this is an improvement, but I highly doubt that it's worse.
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Reformat
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Fix PostgresDriver.delete_all_data()
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Clean up PL/pgSQL code
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* More PL/pgSQL cleanup
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* PL/pgSQL function optimisations
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Ensure compatibility with PostgreSQL 10 and below
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* More/better docstrings for PG functions
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Fix typo in docstring
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Return correct value on toggle()
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Use composite type for PG function parameters
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Fix JSON driver's Config.clear_all()
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Correct description for Mongo tox recipe
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Fix linting errors
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Update dep specification after merging bumpdeps
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Add towncrier entries
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Update from merge
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Mention [postgres] extra in install docs
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Support more connection options and use better defaults
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Actually pass PG env vars in tox
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
* Replace event trigger with manual DELETE queries
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
- When creating a venv, use `python3.7` instead of `python3`
- Remove unnecessary dependency from pyenv pre-requirements on Debian
- Use curl over wget for get-pip on Xenial
Signed-off-by: Toby Harradine <tobyharradine@gmail.com>
This is in anticipation of #2246, although I've written that PR to not break on 3.6, the feature itself is not usable on 3.6. So I think the best way forward is to simply require python 3.7. This also allows devs and cog creators to utilise all of the new features in 3.7, and it also updates the docs so all operating systems will have 3.7 installed.
Signed-off-by: Toby <tobyharradine@gmail.com>
* Correct minimum version
see #2092
While this is needed because of an import just for typing, I see no reason not to bump the minimum version since this is a minor version difference since this is several minor version behind the latest 3.6, and there have been both security and performance improvements since.
That said, we need to be testing on our lowest supported version to ensure we don't have this happen again, right now our tests run on whatever Travis grabs for 3.6, which I assume is 3.6.6, but could be wrong.
* Update other mentions of min version to 3.6.2
