cider--sesman-friendly-session-p: avoid semi-recursive edge case

[vemv]

In the following profiling data, one can see how sesman-current-sessions ends up calling itself:

       14109  89% - ...
       13684  86%  - sesman-current-sessions
       13684  86%   - sesman--friendly-sessions
       13684  86%    - seq-filter
       13684  86%     - seq-map
       13684  86%      - apply
       13684  86%       - #<compiled 0x18438d3d527778b4>
       13684  86%        - mapcar
       13684  86%         - #<compiled -0x1c3c5aba76cb956a>
       13684  86%          - #<compiled 0xa07ea5fb0bf551>
       13684  86%           - sesman-friendly-session-p
       13684  86%            - apply
       13684  86%             - #<compiled 0x12ba488e81173447>
       13684  86%              - cider--sesman-friendly-session-p
       13681  86%               - cider-current-ns
       13681  86%                - cider-current-repl
       13681  86%                 - cider-repls
       13681  86%                  - sesman-current-session
       13681  86%                   - sesman--friendly-sessions
       13681  86%                    - seq-filter
       13681  86%                     - seq-map
       13678  86%                      - apply
       13678  86%                       - #<compiled 0x18438d3d527778b4>
       13674  86%                        - mapcar
       13674  86%                         - #<compiled -0x1c3c5aadc99b3d6a>
       13661  86%                          - #<compiled 0xa07ea5fb0bf551>
       13650  86%                           - sesman-friendly-session-p
       13647  86%                            - apply
       13647  86%                             - #<compiled 0x12ba488e81173447>
       13581  85%                              - cider--sesman-friendly-session-p
       10132  64%                               - file-truename
        7722  48%                                - file-truename
        5634  35%                                 - file-truename
        3819  24%                                  - file-truename
        2233  14%                                   - file-truename
         777   4%                                      file-truename
        1499   9%                               + cider-current-ns
        1143   7%                               + seq-find
         405   2%                               + cider--all-path-translations
          81   0%                               + cider-tramp-prefix
          50   0%                               + seq-filter
          20   0%                                 process-get
           6   0%                                 process-live-p
           5   0%                               + cider-classpath-entries

This PR avoids hitting the (cider-current-repl) call that was responsible for that.

[vemv]

This also leaves all Windows tests green again 🍏

[bbatsov]

I don't quite understand the problem - if we don't need to get the REPL we are in the REPL, right? It seems in the case you can just inline the check for the local var and be done with it.

[vemv]

We want to parse the buffer's ns name from its ns form, if any. And not query the repl just for getting a ns form, because repl code hits Sesman code again.

For that parsing, we want to use this logic:

(or cider-buffer-ns (cider-get-ns-name))

We could inline that, however it would seem nice to keep having a centralized cider-current-ns defun that is able to serve a reasonable variety of use cases.

Hope it's a touch clearer now!

[vemv]

I'll merge this for the sake of not risking propagating more pain to snapshot users.

n.b., the bug was introduced with this addition #3424 so no Changelog entry is needed.

[bbatsov]

Regarding the logic - I figured as much, that's why I have doubts about the solution. It's not really centralized if you have to know when you're in a buffer with the ns present as local var or not (or a ns form) to pass some special param. I think that's a very confusing API for most people, but I guess we can change it later.

[vemv]

Not sure if I can see your point, but obviously, I wouldn't oppose any commit or issue 🍻

[bbatsov]

Not sure if I can see your point, but obviously, I wouldn't oppose any commit or issue 🍻

My point is simply that currently it's very easy to make a similar mistake to the one you've fixed as it's unlikely that someone will figure out what this function behaves differently in different contexts.

[vemv]

Got it. I'll give it a think.