`schema` should support absolute paths

[ankon]

serverless-appsync-plugin/src/get-config.js

Lines 61 to 67 in 18644ea

	const readSchemaFile =
	schemaRelPath => fs.readFileSync(path.join(servicePath, schemaRelPath), { encoding: 'utf8' });
	const schemaContent =
	Array.isArray(config.schema) ?
	mergeTypes(config.schema.map(readSchemaFile)) :
	readSchemaFile(config.schema || 'schema.graphql');

This (somewhat explicitly in the name) assumes that the config.schema would be a relative path, but that is not documented.

1. Is there a particular reason for this requirement, or could this be relaxed?

   diff --git a/src/get-config.js b/src/get-config.js
   index 597e084..17e1764 100644
   --- a/src/get-config.js
   +++ b/src/get-config.js
   @@ -59,7 +59,7 @@ const getConfig = (config, provider, servicePath) => {
      const mappingTemplates = config.mappingTemplates || [];
    
      const readSchemaFile =
   -      schemaRelPath => fs.readFileSync(path.join(servicePath, schemaRelPath), { encoding: 'utf8' });
   +      schemaPath => fs.readFileSync(path.resolve(servicePath, schemaPath), { encoding: 'utf8' });
    
      const schemaContent =
        Array.isArray(config.schema) ?

2. If there is a reason, would you accept a PR to clarify the documentation?

[bboure]

@ankon Yes!. That sounds good.
Usually, the schema is in the same or in a subdirectory of the repository, but I don't see any issue with using Absolute paths.

Feel free to open a PR.

What is your use-case?

In addition to that, I have recently found myself in need of using something like glob expressions, in order to split the schema into several files more easily.

The plugin already supports schema stitching, but it requires specifying all the files one by one.

Something like would be a great improvement:

schema: schemas/*.graphql

Since it touches the same part of the code, is that something you'd like to help with too?

Thanks!!

[ankon]

What is your use-case?

I'm trying to build a serverless plugin that auto-generates part of the configuration for our specific environment. Having such "arbitrary" restrictions is "annoying" (easy to work around, but hitting them always has this "Doh!" feeling)

[globbing]

Interesting one. I probably would consider it part of a different PR/issue, but it doesn't sound particularly hard to do (Famous last words ...)

[sid88in]

Good catch — and it survived into v2. The schema paths were being resolved with
path.join(servicePath, schema), which mangles absolute paths (it prefixes them with the
service dir), so an absolute schema path quietly resolved to nothing.

Fixed in #721 by switching to path.resolve — exactly the change you suggested. Relative
paths still resolve against the service directory; absolute paths are now honored. Added
tests for both.

(The glob part of this issue — schema: schemas/*.graphql — is already supported on v2, so
#NNN just closes the absolute-path gap.)

Thanks @ankon!