[MSOURCES-113] Store source encoding info in a standard file

[jira-importer]

Fred Bricon opened MSOURCES-113 and commented

It'd be nice if IDEs were able to automatically set proper source encoding when attaching sources to project dependencies. 

In eclipse/m2e, we tried to automatically detect encoding but it didn't work out https://git.eclipse.org/r/#/c/116390/

A more reliable way would be to ensure the encoding info was stored inside the source archive, so that IDEs could peek into it and ensure proper settings.

 

---

No further details from MSOURCES-113

[jira-importer]

Karl Heinz Marbaise commented

Hi,
usually the information of the encoding is stored in the pom (project.build.sourceEncoding)....If you download the source package my thought is that you already have the appropriate pom file for the artifact? Do I misunderstand a thing here ?

[jira-importer]

Fred Bricon commented

project.build.sourceEncoding could be defined in a distant parent, so reading that information would require to load the complete maven model, which would be both resource intensive and difficult in a project not depending on Maven tooling.

On the other hand, storing that info directly in the source jar would be super simple to leverage, from both a resource and tooling perspectives

[jira-importer]

Karl Heinz Marbaise commented

Ok thats a good point and yes you have to resolve the whole model..which seemed to be a little bit extreme. I could think of a supplemental file in META-INF/m2e/encoding do you something like this in mind?

[jira-importer]

Fred Bricon commented

I'd rather prefer /IDE/, or /editor/ or /tool/, basically something agnostic rather than something m2e specific.

Then I can lobby the gradle team so they do the same.

[jira-importer]

Karl Heinz Marbaise commented

What about META-INF/IDE/eclipse/encoding would make it open for other IDE's...

[jira-importer]

Fred Bricon commented

encoding really is a cross-IDE issue. I'd rather avoid starting adding IDE specific elements in the jar file until we really need them

[jira-importer]

Karl Heinz Marbaise commented

Ok good point. META-INF/IDE/encoding for the beginning...

[jira-importer]

Fred Bricon commented

Cool, I'll work on the m2e bits once there's something to play with. Thanks Karl Heinz Marbaise

[jira-importer]

Karl Heinz Marbaise commented

Some questions coming into my mind. 

The file itself should have an encoding itself for example UTF-8 ? What about the format of the file? Just something like this (property like):

encoding=UTF-18

Or do you think about something very plain like this:

UTF-18

What about a line ending?

[jira-importer]

Fred Bricon commented

I think encoding=UTF-8 would be best

It gives us the opportunity to do more crazy stuff like defining encoding/some/specific/files=UTF-8 eventually (but probably not).

I have no preference whether to add a line ending or not. I don't think it matters much.

I imagine the encoding file encoding would be UTF-8

[jira-importer]

Michael Osipov commented

I think you guys look for this: https://editorconfig.org/. Please, don't reinvent the wheel.

[jira-importer]

Fred Bricon commented

We could add a META-INF/.editorconfig file in the source jar, if we consider it a de facto standard, else it might be weird to add that tool specific coupling.

[jira-importer]

Michael Osipov commented

Fred Bricon, there is even an Eclipse plugin for that file. If so, it would need to be changed to read from META-INF/.editorconfig. You can already create such a resource without our help, don't you?

[jira-importer]

Fred Bricon commented

The point of this ticket is for generated source jars to be able to advertise their content encoding.

If you have projects depending on junit 3.8.1, which uses ISO-8859-1 encoding for its sources, then assuming you manage to configures encoding for source jars in editorconfig, then each of your project needs to specifically declare a specific encoding for junit 3.8.1. 

So for junit 3.8.1, the ship has sailed obviously, but if an new source jar uses some fancy encoding and the IDE can detect it automatically (as its encoding will be immutable at that point), then the problem is solved without manual intervention.

[jira-importer]

Michael Osipov commented

I do understand its point, but is it that hard to add src/main/resources-filtered/META-INF/.editorconfig and filter that file by the system? You just have to create it once, that's it.