Diversify Raised Exceptions

[Peter Arrenbrecht]

On Wed, Jan 21, 2009 at 7:56 PM, Thomas Burdick
<thomas.burdick at gmail.com> wrote:
> I think you misunderstood a bit.
>
> In the update dialog presented to the user their current working
> directory is given as a highlighted node in a graph. If they choose to
> select another node on another branch and they do not have local
> changes, I see that as being the users definitive choice to switch to
> that particular branch/node as their working directory.
>
> On the other hand, if the update were to tip without regard to
> seeing/knowing where your at currently then the cross-branches error
> is certainly valid.
>
> So no, the users would never be questioning where their local changes
> went, even in the case of having a dialog display the history as a
> graph and them selecting any node other than the one their on
> currently, this is a valid error indeed. The cross branches error is
> not.
>
> This in fact reduces the amount of input required by the user, in
> cases where its senseless to ask "do you want to force this operation"
> I can simply do it for them as its what they've already chosen to do
> and can visually see the results prior to clicking "update".
>
> If it does not make sense to make exception classes then I will end up
> parsing error message strings which is probably much worse.

You could also detect the problematic situations yourself before even
attempting the operation. In the two cases you mentioned this sounds
perfectly feasible.
-parren


> -Tom
>
> On 1/21/09, Matt Mackall <mpm at selenic.com> wrote:
>> On Wed, 2009-01-21 at 11:52 -0600, Thomas Burdick wrote:
>>> certainly, best examples I can think of are the
>>>
>>> update crosses branches and...
>>> update overwrites local changes.
>>>
>>> Those two are the first I've run across. They are util.Abort raised
>>> errors, I'd like them to be unique so I can do something like...
>>>
>>> try:
>>>      dispatch.dispatch(['update', '-r', somerev])
>>> except util.CrossBranches:
>>>      #ignore, this is a pretty useless message, just force it anyways
>>
>> Great. Now you'll have your users wondering where their local changes
>> disappeared to and why they're getting "push adds new remote heads"
>> warnings instead.
>>
>>> except util.OverwritesLocalChanges:
>>>      #pop a message box asking if the user if they'd like to overwrite and
>>> continue or just cancel the operation.
>>>
>>> Does this make sense to do?
>>
>> Not really. We'd end up with about 300 exception classes.
>>
>> --
>> http://selenic.com : development and support for Mercurial and Linux
>>
>>
>>
> _______________________________________________
> Mercurial mailing list
> Mercurial at selenic.com
> http://selenic.com/mailman/listinfo/mercurial
>