Is it even possible to template your dependencies yaml file?

[grepwood]

Hi,

I have a file that starts like so:

options:
  buildHeader: [ "load(\"@io_bazel_rules_scala//scala:scala_import.bzl\", \"scala_import\")" ]
  languages: [ "java", "scala:2.12.10" ]

The project in question is a gargantuan monorepo. Therefore it stands to reason that I would like to change the Scala version in one place and see the change affect literally everything, to save myself work.

I have made the following changes:

1. Add a file in my root directory called "scala_version.bzl" that contains SCALA_VERSION = "2.12.10"
2. I have modified my dependencies yaml like this:

options:
  buildHeader: [ "load(\"@io_bazel_rules_scala//scala:scala_import.bzl\", \"scala_import\")" , "load(\"//:scala_version\", \"SCALA_VERSION\")" ]
  languages: [ "java", "scala:" + SCALA_VERSION ]

This is where the cookie crumbles - I have no idea how to template this file. I tried:

• "scala:" + SCALA_VERSION
• "scala:${SCALA_VERSION}"
• "scala:{SCALA_VERSION}"
• at this point I changed SCALA_VERSION to "scala:2.12.10" and tried "{{SCALA_VERSION}}" because that's a standard in YAML format

Unfortunately, none of those worked. Can someone help please?

[johnynek]

this code uses snakeyaml to parse the yaml.

you can see the extent of the Yaml interop here:

bazel-deps/src/scala/com/github/johnynek/bazel_deps/CirceYaml.scala

Line 14 in b5c7231

new ObjectMapper(new YAMLFactory()).registerModule(CirceJsonModule)

I have never used yaml templating like this. It is possible snakeyaml requires some configurations to work with jackson to support templates.

I am willing to review code with tests to add support, but I will not have time to add support or debug this myself.

I would recommend you use another tool to convert templated yaml into the version you pass to this tool if you don't want to dig into snakeyaml.