Hi David, I think both are valid approaches, with the second one looking more like a workaround, quite similar to what we do with copying to docs/components via Gulp. I think we're all used to committing regenerated files when we forget to commit them initially, could be that we got used to the pain :)
Do you know the timeline for the release of Antora for the first option. I'd go with that if possible, seems like that would have less of an impact on 2.x and 3.0.x branches, so less work there. And you're right, support for symlinks on Windows could bring on more issues[1] zoran [1] https://github.com/git-for-windows/git/wiki/Symbolic-Links On Fri, Mar 27, 2020 at 2:13 AM David Jencks <[email protected]> wrote: > > I’m still not happy about the amount of copying of source adoc files in the > website build. The components adoc files are copied into docs/components > with a gulp script run as part of the docs build, and the copies need to be > checked in. > > I know of two strategies to alleviate this, one available now: > > 1. (works but relies on unreleased Antora features); use an Antora plugin to > gather the files from their locations in the components directory. The > copies in docs/components could be removed. > > 2. replace what are now the “original” files in the components directory with > symlinks to what are now the “copies” in docs/components. Since Antora would > be using the copies that aren’t symlinks in docs/components, that would work > fine. I haven’t tried it, but would expect that the code generation > operations that modify the files based on extracted/generated json files > ought to work on the symlink “copies”. Certainly editors ought to be able to > deal with the symlinked copies. > > I don’t know if (2) would actually work on windows, and have no way to find > out. > > Do either of these seem worth pursuing? > > Thanks > David Jencks > -- Zoran Regvart
