Add cider-load-buffer-and-switch-to-repl-buffer

[jeremyheiler]

@SirSkidmore and I wrote a command C-c M-z to abbreviate:

C-c C-k
C-u C-c C-z

Specifically, the function will load (eval) the current buffer, change the namespace of the REPL buffer to match the currently visited source file, and then switch to the relevant REPL buffer.

We thought about parameterizing the the function to selecting the REPL buffer, but that would have to be done before cider-load-buffer is invoked, and weren't sure how to handle that. While attempting to do it, the message from cider-load-buffer would be in the way unless you know to type something to get rid of it.

[Malabarba]

This behavior would make a lot of sense to me under C-u C-c C-k, given how tightly related it is to C-c C-k.
Code looks good. 👍

[jeremyheiler]

We thought through the command sequence, and we've concluded that the C-c M-z family makes the most sense if we update it to behave just like C-c C-z, but have it always load the buffer.

        C-c C-z =                                               switch to repl buffer
    C-u C-c C-z =                               load namespace, switch to repl buffer
C-u C-u C-c C-z =              prompt for repl, load namespace, switch to repl buffer

        C-c M-z = load buffer,                                  switch to repl buffer
    C-u C-c M-z = load buffer,                  load namespace, switch to repl buffer
C-u C-u C-c M-z = load buffer, prompt for repl, load namespace, switch to repl buffer

Using C-c C-k could work, but isn't as symmetrical with regard to C-c C-z.

        C-c C-k = load buffer
    C-u C-c C-k = load buffer,                  load namespace  switch to repl buffer
C-u C-u C-c C-k = load buffer, prompt for repl, load namespace, switch to repl buffer

[jeremyheiler]

I have updated the function to accept universal arguments a la C-c C-z.

However, when running C-u C-u C-c M-z, the minibuffer shows the Loading %s... message from cider-load-file on top of the prompt from cider-set-relevant-connection. Is it possible to suppress the message, or have the prompt force its way to the front?

[expez]

This behavior would make a lot of sense to me under C-u C-c C-k

👍

This also reduces the cognitive load of yet another keybinding, but just barely :/

[Malabarba]

        C-c M-z = load buffer, switch to repl buffer

If I'm loading a file and switching to the REPL as a single action, I almost certainly want the namespace to be loaded on the REPL.

    C-u C-c M-z = load buffer, load namespace, switch to repl buffer

This looks much more useful than the other one.

Using C-c C-k could work, but isn't as symmetrical with regard to C-c C-z.

It's good to maintain symmetry where we can, but limiting the number of keybinds is important too. Cider is already starting to run out of these. ;-)

I like this proposal:

        C-c C-k = load buffer
    C-u C-c C-k = load buffer,                  load namespace  switch to repl buffer
C-u C-u C-c C-k = load buffer, prompt for repl, load namespace, switch to repl buffer

[jeremyheiler]

If I'm loading a file and switching to the REPL as a single action, I almost certainly want the namespace to be loaded on the REPL.

A counter point: if you're working in user or some other namespace you've already loaded, and want to load the code from a different namespace that it already has required, and switch to the REPL in one action. Also, if you're working in user, theirs the benefit of having clojure.repl automatically referred.

I agree the more common case should be shorter. However, it isn't with C-u C-c C-k. The only benefit is that it's based off an existing command.

When considering C-u C-c C-k, it felt that it overcomplicated cider-load-buffer by adding the REPL and namespace switching. Also, the interactive part would have to consider that buffer is an optional argument. However, that could be easy to do; I'm not sure.

[bbatsov]

Two prefix args?

[jeremyheiler]

See my comment above: #1232 (comment)

[bbatsov]

I'm fine with the idea, but I'm not sure what the default behaviour should be. Seems there might be some point in switching the ns by default, but on the other hand this won't be consistent with the current go-to-repl behaviour.

[Malabarba]

I should probably say that I don't use the REPL buffer all that much, so take my opinion with a grain of salt. :)
In any case, I don't oppose to any of the options.

[bbatsov]

Looking at the code behind C-c C-z, the variant with selecting manually a project looks pretty strange to me. I totally don't remember what was the use-case behind this and I'll probably remove it the moment #1217 is done.

[bbatsov]

OK, so C-u C-u C-c C-z is no more. You can update the PR to reflect this.

[jonpither]

Background on C-u C-u C-z C-z is in #350, the origin living in Hugo's question of "What happens when there are multiple REPLs for a single project".

I would agree with simplifying it away, unless the original concern stands tall.

[bbatsov]

Background on C-u C-u C-z C-z is in #350, the origin living in Hugo's question of "What happens when there are multiple REPLs for a single project".

This was problematic before and will remain to be problematic now as well. Basically you'll just fallback to the default REPL, which seems reasonable to me.

[bbatsov]

One more thing - now all operations are auto-dispatched via the "correct" connection and fallback to the default connection if the connection associated with a project cannot be determined. There's still some room for improvement, but overall people will have to deal with connection switching way less.

[jonpither]

Cool. That was a bit of an elephant in the room in that if Cider knows the 'relevant/correct' connection, then why not always use that info for all operations. It's a pretty big change to make as it's in the core flow. If there's no problems for anyone, then excellent!

[bbatsov]

Well, previous the biggest practical issue was that you could not associate a local project with a connection created with cider-connect. The problem with multiple connections sharing the same project still exists, although I do plan to add some subfolder project support which should solve this to some extent. The only unsolvable problem is what do if you have two connections to different hosts for the same project, but I think that's not pretty common.

[bbatsov]

@jeremyheiler ping

[jeremyheiler]

@bbatsov Sorry for the delay. My son was born shortly after opening this PR. :-)

I should have some time this week to fix it up.

[bbatsov]

@bbatsov Sorry for the delay. My son was born shortly after opening this PR. :-)

Congrats!

[jeremyheiler]

@bbatsov I've pushed some changes, and rebased on from master.

[bbatsov]

👍