On Saturday, 3 March 2018 at 13:17:15 UTC, Martin Nowak wrote:
Why would you need to have that in RAM instead of leaving it to
the db cache layer?
The search "database" right now is an XML file. Keep in mind this
is a static site generator meant to just work offline or when
pushed to github pages and thus avoids dependencies on... well,
anything. Even if the server side search program is running, it
reuses that same xml file as it was the simplest thing that could
possibly work. Reading the file was slow, so I kept it in memory
with a hashtable index. And since it does a pretty good job and
the server hosting Phobos was under no memory pressure anyway,
I've left it to focus on other things for the last year.
So, when I say "optimize that search database" one of the options
would be to actually use a real database layer :)
Where are you hosting?
https://www.hetzner.de/cloud has fairly affordable KVMs
The dub part is on a $5 digital ocean droplet right now. The
phobos one is a spare computer in my house.
That's a fallacy. If one doesn't have to do anything, it
means it's more likely that nothing is done.
Empirical reality disagrees with you. Actually writing the doc
comments is enough effort as it is and making people go through a
12-step program to host it (or deal with ddoc's... quirks instead
of focusing on the actual content) makes it enough of a hassle
that plenty of people just don't bother. That's why so many dub
packages are undocumented now.
Though since I put this up, three different people have already
emailed me asking to delete their ~master caches because they
added doc comments where none existed before and didn't want to
leave the embarrassing empty page up any longer than they had to!
Adding useful documentation is one of the most important
tasks when writing a library and I'd expect any library with
basic quality standards to take care of that.
Yes, and that's why so many users are unimpressed with the
code.dlang.org site as it was. Lots of packages with no online
docs, and those who had them were hard to find, leading them to
believe the whole library was trash.
Well, turns out a bunch of them DO have doc comments, they just
aren't hosted, and among those without doc comments, they are
willing to add them with just a bit of motivation and reduction
of friction.
That's no changing, and it is already making a difference.