Re: thoughts about shelving
From: Julian Foad <julianfoad_at_apache.org>
Date: Sat, 15 Sep 2018 17:12:10 +0100
Stefan Kueng wrote:
Could do! I suppose that would mean less boilerplate code for you (the client) to write.
> * when a shelve operation fails, we have to remove the version with
I'm not entirely happy with the current relationship between a "shelf" object as a container of "shelf version" objects. At the moment a "shelf version" can only exist as a child of a "shelf". It is intentional that you can have a shelf that has no versions. It can still have a log message. (The user might want to name their shelf and write their log message before they start coding, for example, but that isn't a compelling use case.)
I am considering partly inverting the relationship so a single-version shelf would be the primary object and a series-of-versions (for checkpointing and rollback) would be a higher level object that is a collection of single-version shelf objects. What do you think?
> * maybe this one can make it into 1.11: the doc comments don't mention
Thanks. I'll add it.
> * some thoughts [...] you first have to open/create the shelf, save the new
I agree access to a higher level API would be useful, and I'll make a mental note to add that.
At the same time I think it's essential that clients do have access to the lower-level building blocks of the operation, in order to build more interesting and different variants from what I implemented for the command-line client.
So I would like to provide two levels of API. You won't then have to user the lower-level APIs if you don't want to, but any third-party who wants to innovate on the idea will then have the tools to do so.
> in case you're interested, here are screenshots of the shelve and
Thanks for sharing those!
-- - JulianReceived on 2018-09-15 18:12:18 CEST |
This is an archived mail posted to the Subversion Dev mailing list.
This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.