[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: [PATCH] Issue #2188 - [API] svn_client_copy/move should error

From: Philip Martin <philip_at_codematters.co.uk>
Date: 2005-05-14 23:24:53 CEST

"S.Ramaswamy" <ramaswamy@collab.net> writes:

>> The aim is to make "copy url_or_path1 url_or_path2" predictable,
>> i.e. to allow the caller to predict the exact destination URL or path.
>> As far as I can see with your patch the caller still doesn't know what
>> will get created when src_kind is not a directory.
>>
>
> As far as I could make out from the discussions on the lists and the
> tracker, the requirement is for the directory creation behavior of svn
> copy/move to be configurable, when the dst is an existing directory.

I think the aim should be to provide an interface that produces
predictable results.

> Currently if the destination specified by the user is an existing
> directory, and the src is a directory as well, then
> a directory with name "dst + basename(src)" is created - this surprises
> many users, who would much rather have an error thrown or make this
> behavior configurable.

Forget about "users", this is an API so think about "applications".
An application wil be just as "surprised" when it runs

   svn_client_move foo/bar zig/zag

expecting to create a file zig/zag but getting a file zig/zag/bar.
The application has no reliable way to determine which name will get
created.

>> I don't know whether it makes sense to have a flag to retain the old
>> behaviour in the new function. If the "smart" destination logic was
>> relegated to the deprecated functions the new functions might well be
>> simpler.
>
> Not sure what you mean in the last para. Can you be a little more verbose
> :-).

I'm suggesting that a sensible API would always be predictable, so the
current "magic" that tests whether dst is a directory should be
deprecated. Thus only the deprecated functions would have the
checking code, the new functions would simply assume that the
destination does not exist and attempt to create the copy, returning
an error if the some existing item prevents the copy.

I'm suggesting that svn_client_move should be modelled on the C
library function rename(2) rather than the program 'mv'. Similarly
svn_client_copy should be modelled on link(2) rather than 'cp'.

Things get more complicated if we want to support multiple source
targets, we know approximately how it should behave but nobody has
worked out the details, see:

http://svn.haxx.se/dev/archive-2005-04/0075.shtml

-- 
Philip Martin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat May 14 23:26:45 2005

This is an archived mail posted to the Subversion Dev mailing list.