<sphalerite>
clever: this seems like the kind of thing you might know :p is there a way to overwrite argv from bash?
<clever>
sphalerite: shift, and something related
<clever>
cant remember the other thing
<yorick>
gchristensen: would vacuuming the db also work?
<yorick>
also, 2.5m is still not great
<yorick>
maybe we need to see what query it does and EXPLAIN it
atlas_ is now known as nixhubd
nixhubd is now known as atlas_
<MichaelRaskin>
gchristensen: which format took so long to import?
lovesegfault has quit [Quit: WeeChat 2.7]
drakonis has quit [Quit: WeeChat 2.6]
__monty__ has joined #nixos-chat
<elvishjerricco>
Heh, I knew being pessimistic about Time Machine would pay off. Time Machine said it would need to rebuild the backup from scratch because a verification of some kind failed. So I just rolled back the ZFS dataset that the backup is on to the snapshot after the second newest successful TM backup and told TM to try again. TM was like "guess it's ok now, whatever".
<__monty__>
How do you know TM messed up and not the OSX zfs implementation?
<MichaelRaskin>
ed to an address, cross-checking with the posted resident name is not likely.
<MichaelRaskin>
gchristensen: re: your struggle with Russian addresses: postcode gives a single post office, who generally knows the typical problems of the local addresses — so having just the lines with postcode line, city name (usually Wikipedia-checkable), whatever looks like street name, and then all the house/building/apartment numbers separated by dashes (catch: street name could include numbers. too… these should go with t
<MichaelRaskin>
* the mail is delivered to
<MichaelRaskin>
(ouch, the address given to you has numbers in pretty risky order, apartment number is more often _after_ the house number and building-number-inside-the-same-house-number)
<MichaelRaskin>
(anyway, linebreaks matter less as long as postcode is clearly separated and correct)
<__monty__>
Hmm, if anyone had a rigorous system of addresses I'd expect a communist regime to.
<MichaelRaskin>
It's a bit complicated because they still did kind of preserve the inherited logic
<MichaelRaskin>
And in 25 years some people have lost understadning of the system's logic…
<MichaelRaskin>
And then they try to adapt it to the western order of address entries without understanding either system well
<sphalerite>
__monty__: my guess is that elvishjerricco isn't using the OSX zfs implementation.
<MichaelRaskin>
__monty__: The Right Way to write addresses in Russia would be full name on a separate line, large skip, postal code, address from larger to smaller units.
<elvishjerricco>
__monty__, sphalerite: Yea, the ZFS pool is on Linux. TM is backing up to a network share
<__monty__>
MichaelRaskin: Skip?
<MichaelRaskin>
Erm, TeX-speak showing
<__monty__>
Oh
<MichaelRaskin>
The full name line is a separate block with a bit of vertical space before the actual address
drakonis has joined #nixos-chat
<gchristensen>
this query has been running for about 10, taking bets on if it'll be done by the time I get back -- 3-4 hours from now: update builds set project_id = (select projects.id FROM projects WHERE projects.name = builds.project);
<gchristensen>
10h*
drakonis has quit [Quit: WeeChat 2.6]
<__monty__>
How many projects?
<sphalerite>
gchristensen: what odds are you offering? ;)
<gchristensen>
10 projects, __monty__
<__monty__>
**10**?!?
<sphalerite>
__monty__: many builds though probably.
<gchristensen>
110,000,000 builds
<sphalerite>
I think it was several tens of millions?
<sphalerite>
ok, off by a factor of one :D
<sphalerite>
err
<sphalerite>
one order of magnitude
<gchristensen>
the disks have been writing at 200M/s since I started :|
<sphalerite>
hm, I wonder if that would work better with a join?
<sphalerite>
how fast are the disks?
<gchristensen>
SSDs. #postgresql tells me the JOIN'd version and the subselect version almost certainly used the same execution plan, but I haven't checked it
<MichaelRaskin>
Looks like the currently most succesful open successor of XMPP is Matrix, which, of course, has even more overhead in the large-room scenario…
<qyliss>
XMPP doesn't need a successor
<qyliss>
it's going fine
<etu>
qyliss++
* etu
uses xmpp on a mostly daily basis for group chats
* ar
's stopped using xmpp a few years ago.
<ar>
the only contacts i ever had on xmpp were on irc anyway
<MichaelRaskin>
I still run a Prosody instance
<etu>
Hmm, I should set up selfoss
<etu>
I heard good things about it from this channel
<MichaelRaskin>
But I am pretty sure I have spent more effort on trying to get a local sometimes-offline client with a complete chat history including the times when another client device was the only one online — than on getting Matrix running
<MichaelRaskin>
(including adding an actually working version of matrix-recorder to Nixpkgs)
drakonis has joined #nixos-chat
<__monty__>
etu: Yes, I recommend it regularly : ) (Haven't seen others mention it but I'm glad it's become simply a channel fact : D)
<pie_>
Selfoss is a town in southern Iceland
<pie_>
uhhh
<pie_>
sounds recommendable
<pie_>
whats a selfoss
<etu>
__monty__: Well, I kinda have to migrate my freshrss now. I may or may not have had a harddrive crash with an arch install with my freshrss without backup yesterday :p
<etu>
pie_: rss reader
<pie_>
aha
<etu>
__monty__: So I'm setting it up on the new system now which is nixos with at least a zfs mirror with snapshots :p
<__monty__>
Ouch, my condolences.
<etu>
I call it: Early spring cleaning of my feeds.
<gchristensen>
my UPDATE has not yet finished.
<LnL>
:p
<LnL>
gchristensen: oh just read your message about dumping the hydra db
<LnL>
does the host run vacuum?
<gchristensen>
yeah
<LnL>
hmm
<srhb>
My thought was (also? considering LnL's suspicious tone) that reload performance benefits sounds like vacuuming is being interrupted on the original DB.
<gchristensen>
I think the reload improvement was just around having data structured more nearby on disk
<srhb>
Also possible, yeah. :)
<srhb>
But it also does get rewritten as if a vacuum full had been run, which probably isn't ever possible in production due to the exclusive lock...
<gchristensen>
yeah that could certainly be
<LnL>
yeah, you can check when the last vacuum was applied IIRC
<LnL>
but I guess disk could also be a factor
<gchristensen>
I don't suppose any of y'all have secret tricks in know how much progress an UPDATE made
<gchristensen>
not sure if I'm up for another 5 minutes or another 15 hours
<srhb>
Not really.
<gchristensen>
cool cool cool
<gchristensen>
next time I will batch them, 100,000 at a time or something.
<gchristensen>
(a possible outcome here is giving up and just say this migration won't happen)
<gchristensen>
LOG: duration: 55959127.867 ms statement: update builds set project_id = (select projects.id FROM projects WHERE projects.name = builds.project);
<__monty__>
Thank god it's measured in microseconds.
<gchristensen>
thousandths of a microsecond!
<__monty__>
Isn't ms *milli*second?
<gchristensen>
ahh d'oh you're right
<pie_>
it always trips me up that stopwatches have two of the small digits
<__monty__>
Tbh, that *is* plenty.
<gchristensen>
is there a `set -x` but for Perl?
__monty__ has quit [Quit: leaving]
<MichaelRaskin>
If you debug Perl, you have already lost
<gchristensen>
lol
<MichaelRaskin>
Wipe and rewrite
<MichaelRaskin>
Or use strace
<gchristensen>
is there a mysqltuner.pl except for postgresql, and not in perl? :)
hyperfekt has quit [Quit: bouncer quit]
hyperfekt has joined #nixos-chat
hyperfekt has quit [Quit: bouncer quit]
hyperfekt has joined #nixos-chat
<gchristensen>
lol one of these tables has over a billion rows
<leons>
Simply impressed that Postgres can even store that much
<samueldr>
how come?
<samueldr>
is postgresql bad?
<leons>
DBMSs are wonderful, amazing and horrifying pieces of technology
<gchristensen>
it took 90 seconds to count the number of rows
<leons>
No, I love Postgres
<leons>
samueldr: It's just that there is an incredible amount of expertise and work in these systems. I wouldn't want to build such a thing and I'm glad other people do it for me :)
<samueldr>
don't mind me, I was trying to oust some mindless cynicism, but that's not what you seem to be sharing here :)
<samueldr>
it's tiresome that the main answer to any big tech is pretty much always "it bad"
<gchristensen>
yeah
* gchristensen
is guilty of that position
<gchristensen>
wow, I cut down a query's execution time from 7863.594ms to almost nothing
<samueldr>
ooh, exciting, my proof-of-concept is looking good, with a naïve no-optimization run I was able to describe a dependency on files being available for a task to execute
{^_^} has quit [Remote host closed the connection]
<gchristensen>
85082.019 ms -> 26.727 ms
<samueldr>
gchristensen: hydra fast when? ;)
<leons>
samueldr: Mindless cynicism to the max with me, but not in this case, here I'm actually serious ^^
<sphalerite>
yeah I think postgres is one of the good softwares
drakonis has quit [Quit: WeeChat 2.6]
<das_j>
sooo ajs124 is currently using emoji in our config repo…
<ajs124>
intermediate.xml:5499: validity error : xml:id : attribute value opt-helsinki.💩 is not an NCName
<das_j>
does anyone have experience with that?
<das_j>
do we somehow need to enable utf-8 in xsltproc?
<sphalerite>
use 🤮
vika_nezrimaya has joined #nixos-chat
<samueldr>
the bit that translates from options to docbook would need to handle that transparently
<das_j>
samueldr: Is there a tool to do these replacements?