teco-project / teco-code-generators Goto Github PK

These builders are hacky and error-prone. We should switch to SwiftSyntaxBuilder step by step to make the implementation neat and safe.

Testing is important to ensure we don't accidentally regress something or introduces silly bugs.

• Introduces unit tests for basic building blocks and helpers.
• Introduces smoke tests with nicely-crafted pseudo metadata.

Tencent Cloud API model documentation is a mixture of HTML and Markdown, but DocC only supports a subset of Markdown syntax. This means we need to transform HTML tags into DocC-supported Markdown. The tags that may occur are listed as follows:

• a
• b
• br
• code
• del
• div
• dx-alert
• font
• h4
• p
• pre
• strong
• table/tr/td
• ul/li

This is more robust and will solve the issue of teco-service-generator binary being outdated.

Note that we need to do it asynchronously in multiple threads.

When a pull request is merged into the main branch, we should automatically build a new copy of Teco and, if it's different from the latest commit, submit as a PR to https://github.com/teco-project/teco.

When a pull request is sent to TecoCodeGenerators, we should try to build against the latest stable API metadata and generate a diff with Teco to evaluate the update. We should also build the generated library to verify if the patch doesn't break.

We need at least 2 kinds of pull requests:

• A pull request that doesn't affect the way code being generated;
• A pull request that changes the code generation rules or functionalities.

Currently we generates all documentation in English. This looks a little weird with Chinese metadata.

Moreover, we also want to generate an English version of Teco but some metadata is missing (available in Chinese). Try to figure out what's the best practice for localization...

TCCLI documentation is mixed up with MarkDown and HTML, while DocC only supports a small subset of MarkDown.

If possible, we should pre-process the documentation field to make it DocC-compatible.