transbankdevelopers-web

This is a auxiliary repository to help us sync the content changes made on https://github.com/TransbankDevelopers/transbank-developers-docs into the repositories used by Cumbre to build the TransbankDevelopers.cl website.

Most commits here will be automated updates to the latest commit of the master branch of the three submodules:

• slate-tbk (a.k.a "slate"), pointing to our clone of Cumbre's slate machinery used to generate html from markdown sources.

• tbkdev_3.0-public (a.k.a "web"), pointing to our clone of a PHP application provided by Cumbre for us to "preview" the way the markdown (and the code!) will look on the real transbankdevelopers website. It's not really 100% fidelity, but it's close enough. The master branch of that repo is continously deployed by us to http://transbankdevelopers-3-preview.herokuapp.com.

• transbank-developers-docs (a.k.a "docs"), the public repository where the documentation lives and is somewhat readable/browsable as we (ab)use Github's support for markdown. The main win is that anyone can see the changes in the documentation and we can (hopefully) also receive contributions to it without the need to understand all the complexity on the slate side nor the PHP side.

In fact, the reason for this repo is exactly that: automate (if possible) the transition from those simple markdown files published on the TransbankDevelopers organization into the real thing that Cumbre needs to build the website.

How it works

For that, we run the sync.sh script hourly via SemaphoreCI. It basically does the following:

1. Integrate slate changes from two sources:

   • Whatever Cumbre changes in their repository is merged into master (if possible)
   • Whatever is changed on the "docs" repository is copied/translated automatically to the way Cumbre expects it (adding the {{dir}} stuff, renaming every README.md to index.md, etc) and pushed into the transbank-developers-docs branch. Then (if possible) it is merged into master and pushed to Cumbre's repository too.

2. Integrate web changes from two sources:

   • Whatever Cumbre changes in their repository is merged into master (if possible)
   • Whatever is produced by the build-all.sh script on the slate repository is automatically added and pushed to the slate-tbk branch. Then (if possible ) it is merged into master and pushed to Cumbre's repository too.

There are some drawbacks from doing this in a repository that is itself updated by the sync process (we push new changes into the submodules and then we need to keeping submodules up to date in the parent repository) and therefore triggers new unnecesary builds.

But this way provides a lot of traceability: via the CI platform we have full visibility of which commit of each repository was involved on every build. Given that this automation is never going to be perfect and we will have to manually fix conflicts (and bugs on the sync script itself!) from time to time, the traceability/repeatability is very, very nice.