Path: csiph.com!usenet.pasdenom.info!news.redatomik.org!newsfeed.xs4all.nl!newsfeed7.news.xs4all.nl!post.news.xs4all.nl!not-for-mail
MIME-Version: 1.0
In-Reply-To: <87r3o4xfhx.fsf@Equus.decebal.nl>
References: <87380kzb8b.fsf@Equus.decebal.nl> <mailman.711.1437310769.3674.python-list@python.org> <87r3o4xfhx.fsf@Equus.decebal.nl>
Date: Mon, 20 Jul 2015 04:11:51 +1000
Subject: Re: Is this a good way to work with init and exception
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.733.1437329515.3674.python-list@python.org>
Lines: 84
NNTP-Posting-Host: 2001:888:2000:d::a6
Xref: csiph.com comp.lang.python:94144

On Mon, Jul 20, 2015 at 2:46 AM, Cecil Westerhof <Cecil@decebal.nl> wrote:
> On Sunday 19 Jul 2015 14:59 CEST, Chris Angelico wrote:
>
>> Reordering/interleaving your post to respond to different parts
>> together.
>>
>> On Sun, Jul 19, 2015 at 8:35 PM, Cecil Westerhof <Cecil@decebal.nl> wrote:
>>> I am using libturpial to post things on Twitter. But sometimes I
>>> get a ServiceOverCapacity exception. So I wrote the following code.
>>>
>>> ======================================================================
>>> class InitAlreadyDoneError(Exception): pass
>>
>>> Is this the correct way to work user defined exceptions, or should
>>> I also define a default message?
>>
>> I'd start by looking through the exception hierarchy for something
>> appropriate to subclass. In this case, you're basically saying "run
>> init() exactly once, and if you run it a second time, I'll throw
>> back an error", which probably doesn't have any logical match, so
>> directly subclassing Exception would be correct. But you might
>> decide that subclassing ValueError or RuntimeError is more
>> appropriate.
>
> Subclassing ValueError or RuntimeError looks wrong to me.

Sure. Like I said, directly subclassing Exception seemed the most
logical route. Just threw that out there as a possibility.

>> (Side point: It might be a neat courtesy to let people call init
>> again, or maybe a try_init() that won't error out if already
>> initialized.)
>
> I changed it to:
> ========================================================================
> def init(max_tries = 5, wait_time = 60, reinit_allowed = False):
>     global _core
>
>     if (_core != None) and not reinit_allowed:
>         raise InitAlreadyDoneError
> ========================================================================

That works, too!

>>> I use this in the following way:
>>> import twitterDecebal
>>> twitterDecebal.init()
>>>
>>> Because you can not give parameters with an import as far as I can
>>> see. Is this a good way to do this, or is there a better way?
>>
>> Parameterized imports aren't possible, correct. What I'd look at
>> here is a more explicit instantiation. Something like:
>>
>> import twitterDecebal
>> twitter = twitterDecebal.twitterDecebal(5, 60)
>
> I worked with default values, because I thought that would be a good
> idea. I should remove the default values?

No no, the default values are good. I just gave an example that didn't
use them, as that's where you actually need the call. If you're always
going to use the defaults, well, there's not a lot of point having the
function. But if you often use the defaults (or one of them), and
occasionally override it, then what you have is good design.

>> Especially since it's something that does a ton of network
>> operations and all sorts of sleeps and timeouts, I would strongly
>> recommend NOT doing this on import, even if you could. If you don't
>> absolutely _need_ it to be global, it'd be cleanest to make it a
>> class that you construct.
>
> In principal I only mend that before you use the twitter functions you
> need to do the init. (And because of the ton of functions I wanted a
> reinit to be an error.) In my case it is exactly below the import.
> Because I use it in a script and except one situation _core is always
> used. So I thought it to be more clear.

I think it's fine, then. As long as it makes absolutely no sense to
have two separately-initialized twitter connections, and as long as
it's okay for two separate modules to both import this and to then
share state, then what you have is fine.

ChrisA