Path: csiph.com!usenet.pasdenom.info!dedibox.gegeweb.org!gegeweb.eu!nntpfeed.proxad.net!proxad.net!feeder1-2.proxad.net!news.tele.dk!news.tele.dk!small.news.tele.dk!newsgate.cistron.nl!newsgate.news.xs4all.nl!post.news.xs4all.nl!not-for-mail
MIME-Version: 1.0
In-Reply-To: <54539760$0$13009$c3e8da3$5496439d@news.astraweb.com>
References: <5452ADFD.5030801@it.uu.se> <87h9ykphm9.fsf@gmail.com> <mailman.15352.1414757973.18130.python-list@python.org> <54539760$0$13009$c3e8da3$5496439d@news.astraweb.com>
Date: Sat, 1 Nov 2014 01:29:54 +1100
Subject: Re: Saving a file "in the background" -- How?
From: Chris Angelico <rosuav@gmail.com>
Cc: "python-list@python.org" <python-list@python.org>
Content-Type: text/plain; charset=UTF-8
Precedence: list
Newsgroups: comp.lang.python
Message-ID: <mailman.15357.1414765802.18130.python-list@python.org>
Lines: 53
NNTP-Posting-Host: 2001:888:2000:d::a6
Xref: csiph.com comp.lang.python:80019

On Sat, Nov 1, 2014 at 1:06 AM, Steven D'Aprano
<steve+comp.lang.python@pearwood.info> wrote:
> Chris Angelico wrote:
>
>> Sounds like a lot of hassle, and a lot of things that could be done
>> wrongly. Personally, if I need that level of reliability and
>> atomicity, I'd rather push the whole question down to a lower level:
>> maybe commit something to a git repository and push it to a remote
>> server,
>
> How is pushing something to a remote server more reliable than writing to a
> local drive?

It's not, as far as the program's concerned, but when I need
reliability I usually also need replication/backups.

>> or use a PostgreSQL database, or something of that sort. Let
>> someone else have the headaches about "what if AV opens the file". Let
>> someone else worry about how to cope with power failures at arbitrary
>> points in the code.
>
> That "someone else" is still you though. And "use a PostgreSQL database" is
> no solution to the problem "how do I reliably write a 250MB video file?".

If the commit happens, it's saved, and everything else isn't my
problem. And yes, Postgres isn't ideal for a 250MB file, but then,
neither is any sort of autosave good for that. I'm assuming the
content isn't that big.

> Or for that matter, a 2KB text file. Databases are great as databases, but
> they aren't files.

A database can handle a 2KB text file, no problem. That's how most CMSes work.

>> (Though, to be fair, using git for this doesn't
>> fully automate failure handling; what it does is allow you to detect
>> issues on startup, and either roll back ("git checkout -f") or apply
>> ("git commit -a") if it looks okay.)
>
> I don't see how this sort of manual handling is less trouble than using a
> proper, reliable save routine. To start with, what do you mean "detect
> issues on startup" -- startup of what?

Because fundamentally it's impossible to write a reliable save
routine. The best you can do is detect, on subsequent restart of your
program, that there was an incomplete save in a previous session, and
either roll it back or complete it. In some cases that's trivial - if
there's a temporary file, you have the partial save, so you delete the
temporary (rolling back) - or you just ignore it and overwrite, and
hope that it's not going to get stuck anywhere. But it's still the
same broad technique, whichever way you do it.

ChrisA