This is mostly just a follow up to #6342.
`prefer-optional-chaining` was turned on and fixed in every location it
complained in. The transformed function [0] looks expensive from a
glance but from skimming through the replaced sites it doesn't appear
to be ran in any important place, so it should be OK.
The linter improvements are:
- Increase linter performance
- Make `full-lint` and `lint` write to different caches so we
avoid overwriting their caches since they're different configs
- Change husky's hook to `npm run lint` so as to write to the
same cache
- Remove `@typescript-eslint/eslint-plugin-tslint` which is
essentially a wrapper to TSLint because the rules aren't worth
running another linter
- Convert `.eslintrc.json` and `.eslintrc-syntax.json` to two spaces
rather than four tabs to respect PS' `.editorconfig`
- Rename `fulllint` to `full-lint` to ease spelling it
[0] - https://pastie.io/mmtxpf.js (prettified)
`node tools/set-import [version]` can be run to create a
`@pokemon-showdown/sets` package containing sets from Smogon, usage
stats and third party sources. Some notes:
- The build is set up so that `tools/set-import/importer.ts` is
compiled to `tools/set-import/importer.js` - creating a
`.tools-dist/set-import` directory for the sole artifact was
deemed to be overkill
- The sets package is generated such that it can be used easily on
Node or in a browser (where the large set JSON can be loaded
asynchronously)
- Supported tiers/formats are mostly arbitrary - those popular
enough to have a signficant playerbase or analyses on Smogon have
been included, but additional formats can be added based on demand
- Some set information is redundant for ease of use by downstream
users: certain formes are split out and information that could
theoretically be inferred like level/ability/HP IVs are included
to simplify parsing logic and make the sets more immediately
available. This results in what should mostly be negligible
size overhead.
- In a similar vein, display versions of effect names instead of IDs
are used (name -> ID is trivial, ID -> name requires data lookup)
- All sets pass validation, provided certain simple transformations
are applied (eg. reverting `battleOnly` formes like Megas)
This tool has primarily been tested to run on Linux - running on
other platforms is not guaranteed to result in error-free output.
With `-i` in tests/mocha.opts, `npx mocha -g Foo` runs everything
*except* the tests you actually wanted to run (the `-g` overrides
the default `-g`, but unless you specify `--no-invert`, nothing will
override the default `-i`).
Having `test` not just be `mocha` is regrettable, but breaking `npx
mocha -g` or requiring users who want to filter remember to also
pass `--no-invert` is not a good tradeoff.
- Refactor dev-tools/harness.js and dev-tools/smokes.js to separate
out the script/CLI code and implementation code into separate
files.
- Rename 'smoke' to 'exhaustive' ('multi' mode can also be used
for "smoke testing") to better describe its behavior.
- Rewrite the runners in Typescript for type safety.
- Refactor common build utilities into dev-tools/build.js and
introduce the notion of a 'full' build analogous to 'full' tests.
Hotpatching and running `./pokemon-showdown` now automatically run
`./build`. There should now mostly not be any reason you'd want to
manually run `./build`, except if you're invoking tests directly.
In addition, a lot of redundant code has been removed.
I'm not 100% sure this works on Windows, but I'm sure I'll get reports
if anything breaks.
Also move mods/ to data/mods/
This makes PS more monorepo-like. The intent is to further separate
the sim and the server code, but without fully committing to splitting
the repository itself.
We now support `./pokemon-showdown start` in addition to
`./pokemon-showdown`. I'm not clear which I want to be the default
yet.
As far as I can tell, `curly, multi-line, consistent` does everything I
want; there's no reason to keep around a validate-conditionals rule.
Which is probably good, since eslint is deprecating the API for this,
anyway. The nice thing about not relying on deprecated APIs is that now
you can lint PS with `eslint` rather than needing to memorize
command-line switches.
Permessage-deflate is an extensIon that compresses websocket messages with zlib.
SockJS already supports it (indirectly) and the client's load balancer already
inserts the header in the opening handshake when making a WebSocket connection.
Config.wsdeflate makes this optional and allows tweaking the extension's resource
usage.
Modern versions of npm have made it less and less tenable to have
truly optional dependencies, so we'll just bite the bullet and let
people have slightly slower installs in exchange for not having to
deal with npm weirdness.
The new FS module is an abstraction layer over the built-in fs module.
The main reason it exists is because I need an abstraction layer I can
disable writing from. But that'll be in another commit.
Currently, mine is better because:
- paths are always relative to PS's base directory
- Promises (seriously wtf Node Core what are you thinking)
- PS-style API: FS("foo.txt").write("bar") for easier argument order
- mkdirp
This also increases the minimum supported Node version from v6.0 to
v7.7, because we now use async/await. Sorry for the inconvenience!
This also drops the mock-fs-require-fix dependency
mock-fs-require-fix was always kind of a huge hack. It's no longer
necessary, with an FS API that does everything it used to.
This removes a lot of other hacks from test/main.js, which is nice.
`npm install` forces package.json into using the unicode accented-e in
"Pokemon" instead of the escaped format, so committing this prevents the
dirty tree that would otherwise happen every time someone installed
dependencies.
Config.ofe toggles whether or not to write heapdumps if sockets workers
run out of memory, since ofe is an optional dependency but is not
installed by default. nodemailer is now a nonDefaultDependency, and
will complain if it's not installed when Config.crashguardemail is
enabled.
We no longer use it.
Nondeterministic tests are annoying and fuzzers are only useful for
detecting crashes, memory leaks, etc. You can't use them to detect
behavior correctness because that's what your actual code does.
(In other words, fuzzers are useful, but not as part of a regression
test suite.)
- Mocha's new major version fixes the persistent
deprecation warnings on usage of every `npm` command.
- MockFS is locked to 3.10.0 due to an issue affecting
its sandbox capabilities.
https://github.com/tschaub/mock-fs/issues/145
eslint's ignore system first gathers a list of all files, then
decides which ones should be ignored. The "gather a list of all
files" runs out of memory when used on the main server with its
100 million log files, even though the log directory is ignored.
So now we're only passing the files we want linted to eslint.
Hopefully they can fixeslint/eslint#5679 so we can just
`eslint .` again
- Eslint@3 brings some advanced static path analysis to the table.
- Latest version of Mocha doesn`t really bring anything outstanding,
but the caret should cover the next version, which will fix the
current subdep deprecation warnings on `npm install`.
- Mock-FS update brings support for fs.access and fs.accessSync,
which are the Node.js` blessed APIs to replace fs.exists.
An external dependency is kind of overkill when the only
implementation we need is three lines.
This also lets us write a more performant implementation that
omits the parts irrelevant to us.
Currently, only those features are allowed: let/const, classes, and
octal literals. More can be added when needed, as long main Showdown
server has a supporting Node.js's version.
This also accidentally fixes a bug which caused data files to not
be checked for `let`/`const` keyword existence, which would cause
a problem in a web browser.
