ninefx/bs-precis

Bucklescript Implementation for Precis; WIP NPM package for Precis string-prep

bs-precis's Issues

Add all codec tests

Add all tests from RFC8264, 65, 66.

More Tests

It would be nice to generate more tests, to hit more code coverage, and more edge cases. Currently, only the profiles are extensively tested. We would also like to add profile exceptions to be matched correctly, for more thorough testing.

Optimization of Long Switch Statements

The implementation of some of the core unicode functions are generated by the python scripts in ./precis_gen/. They are also either long switch statements, or an optimized b-tree implementation. When trying to use lookup tables, we ran into memory issues. And straight switch statements gave us incredibly long run times. This is the solution I have at the moment, and am open to hearing others.

Examples of B-Tree Optimization

• src/PrecisCodePoints.re
  • fromCodePoint: codePoint => precisType
• src/PrecisCodePoints.re
  • lowerCaseMap: codePoint => codePoint

All switch statements over a certain size should be optimized this way.

Errors in Precis Implementation

Currently we generate tests off of @byllyfish/precis_i18n including their original golden.json. The additional 40,000 tests were added to golden.json. We are missing around 2.5% of the comparisons. At the very least, it would be very helpful to know why, if not patch those issues.

Test Suites: 2 failed, 2 passed, 4 total
Tests:       1025 failed, 42276 passed, 43301 total
Snapshots:   0 total
Time:        101.394s

Documentation

We should begin writing documentation for the library, including precis-gen/.

NPM Package

Once #3 is solved, we would like to package this as an NPM package.