A description may be more useful than any code. Alternatives exist:
gemini://perso.pw/blog/articles/blog-workflow.gmi
For sqlite. To support multiple blogs I'd probably vary the database file used rather than complicating it with a site table. One could make this more complicated with a field to indicate whether an article should be published, or not.
CREATE TABLE blog (
id INTEGER NOT NULL PRIMARY KEY,
epoch INTEGER NOT NULL,
mtime INTEGER NOT NULL DEFAULT 0,
filename TEXT NOT NULL,
subject TEXT NOT NULL,
body TEXT NOT NULL
);
CREATE INDEX blogepoch ON blog (epoch);
CREATE INDEX blogmtime ON blog (mtime);
A CLI tool to make new, list existing, or to edit entries. At the heart it calls out to a text editor that works on a temporary file; the format looks something like an email message,
Epoch 1670907653
Filename yet-another-blog-engine-rewrite
Subject Yet Another Blog Engine Rewrite
# Yet Another Blog Engine Rewrite
...
Known header values get put into the corresponding blog table columns, and remaining text after the header is placed into the body column.
$ blog-about yet-another-blog-engine-rewrite
$ blog-about -l
$ blog-about -e 30
The epoch tracks the "date" of the posting, and the mtime when the article was last edited, which is used to determine what articles need to be updated and can be exposed via RSS or other interfaces.
This script converts the database contents into a static site. Local links that have files for them in ~/tmp get copied into the directory of the blog entry automatically. Improvements could be made to generate monthly or yearly round-ups, to limit the number of entries on on the main index, etc.
Then, rsync stuffs the files up onto the virtual host via a Makefile rule--since OpenBSD ships with make, why not use it?
clean:
git clean $PREVIEW --force -x
sync:
rsync --timeout=15 $PREVIEW --delete --delete-excluded -avz --exclude=.git\* --exclude=/Makefile --exclude=/genblog . gh:/var/gemini/thrig.me/
.PHONY: clean sync
The PREVIEW environment variable can then be used to see what would change without actually doing so by passing a suitable flag for rsync. The "clean" rule is so things like *.core or ktrace.out can be removed. Probably these should get added to an rsync exclude-from file.
$ PREVIEW=-n make sync
...
The "gh" in the "gh:/var/gemini/thrig.me" rsync target is an alias for my "go home" host. This is set in the ~/.ssh/config file.
Host thrig.me vultr gh
Hostname 104.207.156.138
IdentityFile ...
The offline copy is a backup, and can be used locally
to view pages being modified:
$ ls
blog-engine.gmi index.gmi w3m.gmi
$ wv -l blog-engine.gmi
gemini://thrig.me/software/blog-engine.gmi
$ wv -t test -l blog-engine.gmi
gemini://localhost/software/blog-engine.gmi
wv is an alias for the perhaps too complicated ow.c that can be found under
https://thrig.me/src/scripts.git
that uses the mapping
$ grep gemini ~/.ow/dirmap
/var/gemini/(?<rest>.+) gemini://%{rest}
/var/gemini/thrig.me(?:/(?<rest>.+))? gemini://localhost/%{rest} test
Another way to view what a file looks like is via the editor configuration, here for vi (there are two ^M after the % in the actual configuration file):
map \t :!amfora %
which merely maps backslash-t (for "try it out") to running a gemini browser on the filename. This however will not have the full features of a gemini server, autoindex in particular. One could also make a mapping of
map \y :!wv -t
to direct the URL of the file to the server on the localhost interface. But that will take longer, as instead of using a file on disk it must go though a gemini server. This is a custom version of vi, forked from the base version found on OpenBSD, and is delightfully lacking in features, and thus very fast.
How about some other software?
tags #blog #sqlite #perl