Ok, the wait is over!
I've finally "finished" work on redo-blog2, which became the successor of GitSite.
I originally wrote GitSite because I liked the concept of some set of static files where metadata was stored on the FileSystem and in Git. When I published it, I was happy with where I'd gotten. There were things I wanted it to do that I hadn't gotten around to, but that was alright. At first.
Very quickly I started wanting to do things with it that it wasn't built around. Things like having articles and other things together, without everything being an article.
But I did nothing. Rather than fix the issues I just held off publishing anything.
One of the other issues with GitSite was that it was a big script that ran from top to bottom. It wasn't super easy to handle. Also, it built everything all the time.
Around this time I also saw redo, which is a build system (like Make) that is based off shell scripts. It's less like Make, and more like a collection of tools you can string together to make your own build systems. Kind of.
Anyway, I'd been looking for something to play around with redo on, and I had a site built out of files that I wanted a more modular and efficient build system for. Bam!
So I thought about it for a bit and then did nothing. Stephen eventually wrote some scripts and emailed me a tarball of them called redo-blog based on things I'd said I'd wanted but never done. I liked it some, but it wasn't quite what I wanted.
The biggest breakthrough in redo-blog, which solved an issue I was blocked by on the imaginary thing in my head, was storing intermediate data in YAML. It allowed redo to build from one file to another while keeping a bunch of metadata around also.
So, then a long time passed where I moped around not publishing things.
Finally I sat down, interested again, and wrote the beginnings of redo-blog2. I built it from the ground up to satisfy some of the issues I'd had with GitSite. Namely, the base organizational structure in redo-blog2 is "tags" Without tags, they don't get deployed or built. Articles can have "article" tags, stories can have "story" tags, in then I end up with "/article/blah.html" and "/story/blee.html". I was happy.
Then I did nothing for a while. Yesterday I finally got around to actually finishing up the final needs I had for the system and moving this site over to it.
Redo-Blog2: A description
Ok, so.
First, one needs redo. Or at the very least minimal/do from that repo. Minimal/do is a script that builds everything every time, so it's not super efficient to dev with, but it does work the same as a more full redo system.
Next, all my pages are in files like blah.mime. This is because they are in a pseudo-mime-format. Here's an example:
Title: First Article Content-Type: text/plain Tag: article Tag: test This is the body of the article here. Isn't it exciting that I can put text here?
So, the first part of the file is Headers. These can be arbitrary data, but obviously if a script later doesn't use them, they go away.
Then there's a blank line. Everything else is the page contents.
The data path for redo-blog2 goes as follows:
- First, blah.mime becomes blah.augmented. This involves adding auto-generated Headers to the file to augment the manual ones the author wrote. In my case, this is where Git is probed for info about the file.
- blah.augmented then uses a typeconvert script to turn the body from whatever Content-Type was specified into html. If no type is given it assumes text/plain. This is then recombined with the headers into blah.converted
- This is then run through a tagtemplate for every tag it is a member of to generate a different html file of the form blah.tag.html The reason for this is that one can have a different output structure for the "story" tag, or the "video" tag or whatever. If one doesn't have a special tagtemplate it will make one automatically out of the default.
- Finally, the deploy step moves blah.tag.html into tag/blah.html
There's a lot of other stuff it does around this, but that's the main concept.
Using It
For the author, adding a new article involves making a new mime file and then running "redo deploy". It will only rebuild what it has to (or maybe slightly more) and then move it into place.
The main places where someone setting this up has to touch is default.tagtemplate.do, default.tagindex.do, and default.tagfeed.do.
Into default.tagfeed.do one mostly just has to change the data like "Author Name", Email address, Website, etc.
default.tagindex.do is the file that constructs the files that go into tag/index.html and includes a listing of all items with that tag. You must therefore change that file to include the structure you want.
default.tagtemplate.do is slightly more gross. It creates the files that create the blah.tag.html files. Other than that, though, you just have to change the markup in it to be the one you want.
It's feasible that the only changes you'd need to make are removing my data. The markup is (I think) pretty standard and most of what I do comes with styling.
Conclusion
So, so far, I'm quite happy with this system. I don't yet know of things that will keep me from using it.
I'm sure they're out there, but I haven't come across them yet, and that's pretty good.
The biggest issue, in general, is that redo-blog2 isn't a script, like GitSite was. It's more a concept, or idea.
The scripts in http://github.com/psycotica0/redo-blog2 are what I based the scripts in http://github.com/psycotica0/Site off, but in the end they are disjoint. They're related in concept, but there's no connection between them.
So, if you want to adapt it to work with your site, you can start with either this site, or the redo-blog2 repo, but it's likely there won't be a lot of cross pollination.
I'm alright with that for now.