#cargo #git #tasks #repo

app cargo-pando

Perform tasks concurrently over multiple copies of your repo, over multiple toolchains and/or git revisions

4 releases (2 breaking)

0.3.0 Nov 7, 2019
0.2.0 Feb 9, 2019
0.1.1 Jan 8, 2019
0.1.0 Jan 8, 2019

#192 in Cargo plugins

MPL-2.0 license

629 lines

cargo-pando   Latest Version Rustc Version 1.34.2+ Build Status

Perform tasks concurrently over multiple copies of your repo.

  • test your code against multiple rust releases in parallel with a snazzy progress bar
  • test the index / stage of your repo to validate incremental changes
  • do both of the above at the same time
  • run some other custom command across any of the above checkouts

The name pando comes from the clonal colony of "multiple" trees that are actually one single organism. It is latin for "I spread out".


HERE BE DRAGONS. This extension is in the early stages of development and may cause data loss or worse. Only use if you're very comfortable with git and have backups.

There may also be backwards incompatible changes for each version.


Will be easily installable from crates.io once it's more mature.

git clone (repo url here)
cd cargo-pando
cargo install --path .


git pull origin master
cargo install --path . --force

How it Works

  1. Figure out what toolchains to run against, either from the CLI, .travis.yml, or just using all the installed ones.
  2. Create a copy of the repo's code in target/pando per toolchain, e.g. target/pando/1.31.0. Note that this is destructive. The configured target dir in your cargo config is respected.
  3. Run rustup run TOOLCHAIN_HERE cargo test or some other action in each copy of the repo-- in parallel. For example, cargo +1.31.0 test in target/pando/1.31.0/working_dir.

Output is logged to target/pando/TOOLCHAIN_HERE/output, and each line is printed next to the progress bar for the checkout.


If your tests rely on external resources, keep in mind they won't be in the expected location.

If there are exclusive resources, you'll have to synchronize access yourself.

More parallelism doesn't always make things faster, especially since compilation can be IO intensive as well as CPU intensive.


See cargo pando help for more details.

Test the working directory against the toolchains listed in .travis.yml:

cargo pando test

Test against every installed toolchain except the default, limiting it to 2 cargo tests at any given time:

cargo pando --all test -j 2 

Test each specified toolchain, but only doc tests:

cargo pando -t stable -t beta test -- --doc

If you want to run a single command across all of the checkouts at once, use print, cut, and xargs:

cargo pando print | cut -f 2 | xargs ls

Run an arbitrary command against each checkout, substituting the name of the toolchain where applicable:

cargo pando each echo the toolchain '{}' has been copied

If the command does not lend itself well to the single line given by the progress bars, xargs can help again:

cargo pando print | cut -f 1 | xargs -L 1 -P 2 echo the toolchain is


Test the given toolchain against the index (stage) of your repo. Useful if you're incrementally adding changes to a commit and you want to check that your work in progress still works.

cargo pando --index -t stable test

Handy related commands

See how much space the pando directory is taking up:

du -sh target/pando

Get rid of it!

rm -rf target/pando

Bugs / Open Questions

  • How can we know what files need to be copied over?
  • How can we know when / what files to delete from checkouts?
  • How can we tell the canonical name of a toolchain, to not duplicate between their travis representation and the --all representation?



  • get target from cargo metadata instead of assuming
  • add support for other exec targets
    • print
    • cargo
    • build
    • cmdeach / cmdall (have it print and consume it via shell / xargs!)
      • document that
  • heck, document everything
  • [ ] document using cargo aliases to help with common sub-commands
  • [ ] document helpful env vars

(I can't remember what those last two were about. Oh well.)


  • start writing tests
  • call for (and get) feedback
  • figure out what the earliest compatible rust version is
  • support allowing failures from travis.yml


  • blog post
  • pointing progress bars to stdout
  • answer: would one ever need more than Cargo files, test, and src? (build.rs maybe, and then more?)


  • invoke subtasks with --message-format=json for better output information?
  • determine number of steps for task from dependency list?
  • colorize / emojify output


  • tmux integration (might have to refactor when output is created, etc.)
  • Other toolchain selection / isolation mechanisms?
    • Docker?
    • Can we arbitrarily support this? Might not be worth it.
  • vastly cleaning up actions
  • make print take args to select field deliniation (whitespace, null, ASCII tabular)


~360K SLoC