Slow loading when path subdirectory doesn't have git initialized

[arielcorte]

After I installed the tool and configured the path to my coding directory, loading the sessions took a few seconds.

I started to indagate into the problem, by copying one by one the subdirectories to another path. I found that when I added a subdirectory without an initialized git repo, the time was increased by a few seconds (instead of milliseconds).

I tried configuring --max-depths 1 with no noticeable effect.

The workaround that solved the loading time for me was to make a script to run git init in each subdirectory.

[jrmoulton]

I'm not sure what is going on in your specific case. I have subdirectories without initialized git repos and don't have any delay. Do you have any other specific information?

[arielcorte]

I don't know what info might help you here.

I have my coding directory with 107 repos (the one I initialized git in all of them) and works faster than a test directory in which I added 3 non initialized git subdirectories.

➜  ~ tms config -p coding 
Configuration has been stored
➜  ~ time tms
tms  0.02s user 0.08s system 13% cpu 0.695 total
➜  ~ tms config -p tms_test 
Configuration has been stored
➜  ~ time tms
tms  0.03s user 0.21s system 28% cpu 0.830 total
➜  ~ ls tms_test 
no-git2  non-git  not-git
➜  ~ ls tms_test/not-git 
➜  ~ ls tms_test/non-git 
112A_cf.cpp    1630_cses.cpp  1661_cses.cpp  1955B_cf.cpp  2004A_cf.cpp  230A_cf.cpp  2428_cses_copy.cpp  codeforces  cses              template.cpp
1399A_cf.cpp   1643_cses.cpp  1662_cses.cpp  1956A_cf.cpp  2004B_cf.cpp  231A.py      2428_cses.cpp       contest1    glistering_melon  test_template.cpp
1629_cses.cpp  1660_cses.cpp  166A_cf.cpp    1956B_cf.cpp  2004C_cf.cpp  231.cpp      a.out               contest2    printing          tmp.cpp
➜  ~ cd coding/
zsh: do you wish to see all 107 possibilities (22 lines)? 

coding directory with 107 subdirectories:

➜  ~ tms config -p coding 
Configuration has been stored
➜  ~ time tms
tms  0.02s user 0.08s system 15% cpu 0.619 total
➜  ~ time tms
tms  0.02s user 0.08s system 18% cpu 0.531 total
➜  ~ time tms
tms  0.02s user 0.08s system 21% cpu 0.442 total

tms_test directory with 3 non initialized git directories:

➜  ~ tms config -p tms_test 
Configuration has been stored
➜  ~ time tms
tms  0.03s user 0.21s system 30% cpu 0.782 total
➜  ~ time tms
tms  0.03s user 0.21s system 32% cpu 0.742 total
➜  ~ time tms
tms  0.02s user 0.22s system 29% cpu 0.814 total

I think it is important to also note that the subdirectories of tms_test are directly copied from coding but removed the .git repository

[arielcorte]

I've also noticed that python directories are also causing a bit of a delay with its .venv. Here's the time of tms after deleting it from tms_test/no-git2/deep_learning/.venv:

➜  ~ tms config -p tms_test
Configuration has been stored
➜  ~ time tms
tms  0.03s user 0.15s system 24% cpu 0.712 total
➜  ~ time tms
tms  0.02s user 0.15s system 25% cpu 0.676 total
➜  ~ time tms
tms  0.02s user 0.15s system 27% cpu 0.623 total

which is still greater than the coding directory

[arielcorte]

After deleting the other python venv directory in tms_test/non-git/.../venv, the times make more sense:

➜  ~ tms config -p tms_test
Configuration has been stored
➜  ~ time tms
tms  0.01s user 0.02s system 2% cpu 0.870 total
➜  ~ time tms
tms  0.01s user 0.01s system 7% cpu 0.328 total
➜  ~ time tms
tms  0.01s user 0.01s system 3% cpu 0.638 total

All the tests were done with --max-depth = 1 so shouldn't tms skip the venv directories?

[tstachl]

Not sure if that's the same issue but there definitely is an issue:

❯ time tms

________________________________________________________
Executed in   36.21 secs    fish           external
   usr time    2.41 secs    0.36 millis    2.41 secs
   sys time   15.66 secs    1.61 millis   15.66 secs

[junglerobba]

@tstachl I'm afraid this needs more information, like which version you're running, and possibly your configuration (tms config list). We've also switched to a different git library since the last release, so maybe try running the latest commit from main and see if that helps, if you're not already?
Might also be time for another release...

[tstachl]

Hi @junglerobba, thanks for the quick reply. Here is the requested information:

❯ nix shell nixpkgs#tmux-sessionizer
Welcome to fish, the friendly interactive shell
Type help for instructions on how to use fish
❯ tms --version
tmux-sessionizer 0.4.5
❯ time tms

________________________________________________________
Executed in   41.90 secs    fish           external
   usr time    2.42 secs    0.31 millis    2.42 secs
   sys time   16.83 secs    3.12 millis   16.83 secs

❯ tms config list
display_full_path = false
search_submodules = false
recursive_submodules = false
switch_filter_unknown = false
session_sort_order = "Alphabetical"
excluded_dirs = []
sessions = []
bookmarks = []
clone_repo_switch = "Always"

[[search_dirs]]
path = "~/workspace"
depth = 10

[picker_colors]
highlight_color = "LightBlue"
highlight_text_color = "Black"
border_color = "DarkGray"
info_color = "LightYellow"
prompt_color = "LightGreen"

[shortcuts]
backspace = "backspace"
ctrl-a = "move_to_line_start"
ctrl-c = "cancel"
ctrl-d = "delete"
ctrl-e = "move_to_line_end"
ctrl-j = "move_down"
ctrl-k = "move_up"
ctrl-n = "move_down"
ctrl-p = "move_up"
ctrl-u = "delete_to_line_start"
ctrl-w = "delete_word"
delete = "delete"
down = "move_down"
enter = "confirm"
esc = "cancel"
left = "cursor_left"
right = "cursor_right"
up = "move_up"

[session_configs]

[marks]

I actually ended up packaging the original tmux-sessionizer and am now looking into Sesh.

[junglerobba]

@tstachl Two things:

1. depth = 10 might be an issue if there are any directories that are not git repos, because then those will be searched recursively. If all of your projects are on the top level of ~/workspace I'd recommend setting depth to 1, but simply a lower number than 10 will already help a lot.
2. There have been a bunch of improvements (including performance) since the latest release, and this repo provides a flake package, since you're using nix already. So until the next release is tagged (which I guess should be soon-ish), you could try using the latest build from source.

[tstachl]

@junglerobba version 0.5.0 and depth = 3 is a lot faster:

❯ nix shell github:jrmoulton/tmux-sessionizer#tmux-sessionizer
Welcome to fish, the friendly interactive shell
Type help for instructions on how to use fish
❯ tms --version
tmux-sessionizer 0.5.0
❯ time tms

________________________________________________________
Executed in    4.04 secs      fish           external
   usr time   47.92 millis    0.35 millis   47.56 millis
   sys time  111.88 millis    1.37 millis  110.51 millis