feat(disableEval): add static parsers

[thomasgauvin]

I'm hoping to help push for getting the interpreting parser available (this makes it possible to use mysql2 as an alternative to mysql on Cloudflare Workers).

I opened this as a new PR because I have some bandwidth to address further comments in the PR if they come up. This code merely applies the recommended fixes as noted by @shiyuhang0 to the original PR by @sidorares #2099 (the original commit history is maintained and the branch was rebased onto master).

This could also be created as a PR to deoptimise-large-rows branch, however the maintainers recommend moving forward.

[codecov]

Codecov Report

Attention: Patch coverage is 97.08738% with 12 lines in your changes missing coverage. Please review.

Project coverage is 88.70%. Comparing base (f2cc820) to head (61ca972).

Files with missing lines	Patch %	Lines
lib/parsers/static_binary_parser.js	95.26%	10 Missing ⚠️
lib/commands/query.js	95.23%	1 Missing ⚠️
lib/connection_config.js	94.73%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #3365      +/-   ##
==========================================
+ Coverage   88.45%   88.70%   +0.25%     
==========================================
  Files          83       85       +2     
  Lines       13067    13446     +379     
  Branches     1393     1531     +138     
==========================================
+ Hits        11558    11927     +369     
- Misses       1509     1519      +10     

Flag	Coverage Δ
compression-0	88.70% <97.08%> (+0.25%)	⬆️
compression-1	88.70% <97.08%> (+0.25%)	⬆️
static-parser-0	86.28% <18.20%> (?)
static-parser-1	87.09% <96.84%> (?)
tls-0	88.11% <97.08%> (+0.26%)	⬆️
tls-1	88.47% <97.08%> (+0.25%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[wellwelwel]

Thanks, @thomasgauvin! And sorry, january was a hectic month and I didn't really pay attention to your comment in #2289 (comment).

lib/parsers/static_text_parser.js 11.26% 126 Missing ⚠️

I'll add the label needs tests, but I don't intend to block this PR progress because of this. Probably tomorrow I'll go through the tests that currently use evaluation and adapt them for the new parser.

Perhaps a simple approach with parameterized tests can make things easier to avoid duplicate tests.

[wellwelwel]

@thomasgauvin, I added some security points that were resolved after PR #2099 was created.

If it's not a problem, can I commit directly to your branch (including the typeCast / nestTables question and to add the tests I mentioned yesterday)?

[wellwelwel]

Also, in #2289 it was decided to use disableEval instead of useStaticParser (I'm happy with both, to be honest):

• Adapt to Cloudflare Workers environment #2289 (comment)

[thomasgauvin]

@thomasgauvin, I added some security points that were resolved after PR #2099 was created.

If it's not a problem, can I commit directly to your branch (including the typeCast / nestTables question and to add the tests I mentioned yesterday)?

Yes, please feel free to commit anything to the branch directly. I've added you as collaborator on the fork so you should have commit access.

[thomasgauvin]

Also, in #2289 it was decided to use disableEval instead of useStaticParser (I'm happy with both, to be honest):

• Adapt to Cloudflare Workers environment #2289 (comment)

Missed that, will adjust

[wellwelwel]

@sidorares, as the query (and eventually execute) method is one of the most critical in the entire project, I chose not to create an exclusive test for the static parser, but rather to test the entire test suite against this option, including all Node.js, Bun, and Deno versions.

@thomasgauvin, I won't be committing until Saturday. Feel free to continue or wait 🙋🏻‍♂️

• For the reasons in my comment above, naturally all CI using this option will fail at the moment.
• I'll wait to add this variation to the coverage CI when all the tests have passed.

[wellwelwel]

Just inspired on an Sunday dawn 🌌

Static binary parser implementation

I've implemented the new binary static parser preserving the code as faithful as possible to the original, except for two things:

Unified/simplified the typeCast source logic (globally or via execute method):

const typeCast =
  options.typeCast !== undefined ? options.typeCast : config.typeCast;

A minimal memory improvement, by pre-allocating the length of null bitmaps instead of []:

const nullBitmapLength = Math.floor((fields.length + 7 + 2) / 8);
const nullBitmaskBytes = new Array(nullBitmapLength); // ⬅️

for (let i = 0; i < nullBitmapLength; i++) {
  nullBitmaskBytes[i] = packet.readInt8();
}

---

Testing

As previously mentioned, there are no new tests due to new parsers. Instead, every single test, including the matrixes (CI), platforms and their respective versions (Node.js, Bun, Deno), as well as the entire test suite are tested against the new parsers globally.

Note

Although this is a rather intensive approach, it ensures that a modification in one parser will cause failures if the other parser is not updated accordingly, as well as being easier to maintain in the long term by allowing the same test to be run dynamically for each variation.

---

Coverage

As coverage is tested on different matrixes (CI), naturally the percentage when testing only standard or static parsers will reduce the total coverage count. For this reason, I adapted the limits to support this difference.

---

As far as I'm concerned, it's ready to merge, so please feel free to review 🙋🏻‍♂️