new Makefile, copy static files from other directory

[akoehn]

Introduces a basic Makefile, as discussed in #316.

Files previously residing in hugo/ are now kept in hugo-src and copied
as part of the build process, so that rm -rf hugo cleans everything.

The Makefile consists of a set of phony targets, each building a part
of the site.

Relevant targets:

• site: creates the site
• all: cleans output, checks validity of XML and creates the site
• serve: starts a web server on localhost:8000 to browse the site
• upload: rsyncs the site to the aclweb webserver

This means: make all serve will rebuild the site from scratch and start a webserver,
make all upload will rebuild the site and upload it to the official web server.

Comments welcome!

[mbollmann]

Thanks for the PR! I can see that this is already quite handy for only performing specific steps of the build chain, without having to know or look up the exact commands to achieve them.

Some comments:

1. Your commit duplicates the files in hugo-src/. I believe when they're moved (i.e., added in hugo-src and deleted in hugo) in one commit, it's possible to keep tracking the change history of the files (with git blame). I'm not sure what happens if this is stretched out over several commits. It would be nice to make sure that history doesn't get lost, if possible.

2. Before we decide to merge this, we should also make sure to update the documentation, and probably the CI build script as well. (Just leaving this here as a reminder.)

3. Just curious: Do you think it's possible to replace some of the phony targets with concrete ones? Or is that infeasible due to the huge number of (partly dynamically generated) files?

[akoehn]

Re 1.: That is an error, I was obviously too tied yesterday; I thought I did a git mv.

Yes, please consider this a work in progress. I just wanted early feedback.

That would certainly be possible, the dependencies can be provided by an external program. This would however mean that the python scripts can be instructed to only (re-)generate a subset of the data. I had a short look at that and we could discuss it later.

[mbollmann]

An addendum: I tried the python -m http.server command to browse the generated site, and it didn't really work, since the way Hugo is configured right now the site expects to live under /anthology on the server...

[mjpost]

Thanks for this!

It's a minor aesthetic detail that I could easily adapt to, but I dislike having a hugo-src directory instead of just hugo. Would it be difficult to keep the current structure and generate to hugo-build or build/hugo or something like that? (It would also make the PR smaller.)

[akoehn]

I have updated the pull request:

• files in hugo are still in that place (although hugo does not seem to be a fitting name anymore)
• the site is built into the build directory.
• READMEs updated
• travis.yml updated/
• make serve works

[akoehn]

Oh, I forgot to mention, as I stated this PR to be preliminary before: Please have a look and if it looks good to you (@mjpost, @mbollmann), I'll merge it.

[mjpost]

This is great. I had to make one small change since the sed -i invocation is not platform-neutral, sadly, and broke on OS X.

(I think you saw I had you added as a contributor. Feel free to use local branches now, which make updating easier)

[mjpost]

I merged it locally, since I couldn't update your branch, but also Github wouldn't let me create a new PR since it tied the branch I created to your remote PR, somehow.

Sorry to deny you the pleasure of closure!