option for integration with tmux for more convenient switching / workflow

[strazto]

Hi, I use tmux a lot in my terminal workflows, and a nice tmux setup can really help the terminal feel more comfortable than an IDE.

I wonder if you'd be interested in adding some optional tmux integration to this package?

Something along the lines of:

• if options(rvisidata.use_tmux) is true
  • Detect if we are currently in a tmux session

    Sys.getenv("TMUX")
    #> [1] "/tmp/tmux-1000/default,31069,1"

  • If not, use the standard vd command
  • Otherwise:
    • Split to a new pane in the tmux window
    • Run vd on our temporary csv in that pane
    • Return focus to the R pane (? probably most useful)

Scripting tmux to perform a sequence of commands is not too hard.

I guess that script could go in inst, to provide a decent default tmux environment for viewing the data.

There's not really a single "best" workspace for mixing between data exploration and interactive console use, so I imagine that in some instances it will be more useful to open vd in a new tmux window, etc.

If that's the case, maybe allowing users to swap in their own scripts may helpful at some stage.

I'll fork and PR for now, and let me know what you think!

[strazto]

I've pretty much implemented this, except that neither system nor system2 are clever enough to know to syncronously wait for vd to execute and return in a different tmux pane, as now it is not waiting for vd to return before continuing execution, and tmux split-window, etc don't wait before return.

The result of this is that the temp file tends to be deleted before vd can get to it.

If we omit file.remove(temp_file) at the end of vd, everything works correctly, but the file is not immediately cleaned up.

To me, this is not a huge issue, as the R session will still clean up temporary files automatically when it terminates, but you might disagree.

What are your thoughts?

[paulklemm]

Sounds really good. I'm an avid TMUX user myself, so this is a huge help. Merged the PR to master branch.

[paulklemm]

I get an error executing the shell script even though I'm already in a TMUX session. Will have to look into what's happening here.

/beegfs/scratch/bruening_scratch/pklemm/R/user_lib/rvisidata/tmux_pane.sh: line 5: tmux: command not found
Warning message:
In system2(vd_cmd, args = temp_file) : error in running command

[strazto]

Hm - funny What version of tmux are you running?

[strazto]

Not sure when split-window was added to tmux

[paulklemm]

I'm running tmux 2.6. But the error indicates that the script cannot locate tmux, which is odd. Will look into this when I have time.

[strazto]

Very odd, Maybe its a difference in where and when you source your environment?

If the bin dir that gives you tmux isn't added to PATH in a place that's sourced by R, maybe that could have something to do with it?

Its working on ubuntu, and on my work's Centos HPC env, though.

[strazto]

https://stackoverflow.com/questions/26065564/why-cant-i-call-system-functions-in-r-that-clearly-work-in-my-terminal SO thread describing a similar problem

IMO your best bet at a quick fix is to symlink /usr/bin/tmux to whatever which tmux spits out

[paulklemm]

❯ which tmux
/usr/bin/tmux

Well, it's already installed to /usr/bin/tmux. What a weird error.

[paulklemm]

Nevermind, I'm stupid. I'm running R on a HPC cluster using Singularity. R searches in /usr/bin of the Singularity container, not the actual host system. I guess I can fix that by adding tmux to my Singularity image.

[strazto]

Closed by #4