Skip to main content

Command-line script to re-run the given command when files change.

Project description

In most circumstances, this project is superceded by https://github.com/tartley/rerun2 a ten-line Bash script which uses inotify instead of polling, and gracefully handles commands containing shell aliases and the like. However, rerun2: * Won’t work outside of Linuxland. * Doesn’t see filesystem events on mounts in a VM. * Doesn’t take any command-line args to modify behaviour. so this original ‘rerun’ still has its uses.

Install

pip install rerun

Usage

rerun [--help|-h] [--verbose|-v] [--ignore|-i=<file>] [--version] <command>

Where:

<command>           Command to execute as a single arg, i.e. put it inside
                    double quotes, or else spaces and other special
                    characters should be escaped.
--help|-h           Show this help message and exit.
--ignore|-i=<file>  File or directory to ignore. Any directories of the
                    given name (and their subdirs) are excluded from the
                    search for changed files. Any modification to files of
                    the given name are ignored. The given value is
                    compared to basenames, so for example, "--ignore=def"
                    will skip the contents of directory "./abc/def/" and
                    will ignore file "./ghi/def". Can be specified multiple
                    times.
--interactive, -I   Run the command in an interactive shell. This allows
                    the use of shell aliases and functions, but is slower,
                    less reliable and noisier on stdout/stderr, because it
                    sources ~/.bashrc and the like before running the
                    command. Not available on Windows.
--verbose|-v        Display the names of changed files before the command
                    output.
--version           Show version number and exit.

Example

rerun --verbose --ignore=myoutputdir "python -m unittest mymodule"

This will run your tests whenever you save your source code in the current dir or its subdirectories, but it won’t rerun the tests a second time when .pyo files get updated as a result of executing the tests, nor when our program writes to myoutputdir.

This is handy for seeing the new test results in another console window after you hit ‘save’ in your editor, without having to change window focus.

Description

Rerun detects changes to files by polling file modification times once per second. It looks in the current directory and all its subdirectories. On detecting any changes, it clears the terminal and reruns the given command.

It always ignores directories called .svn, .git, .hg, .bzr, build and dist. Additions to this list can be given using –ignore.

It always ignores files ending with .pyc or .pyo. This isn’t currently user-overrideable.

While polling sounds sub-optimal, I’ve yet to encounter a project large enough that rerun’s resource usage was even noticeable. (Plus, see discussion of ‘watchdog’ below.)

Dependencies

Tested on MacOSX, Ubuntu, WindowsXP, Windows 7.

Tested on Python 2.6, 2.7 and 3.4.

Installing using ‘pip’ will also install any other Python packages that are required. (see setup.py ‘install_requires’)

Hacking

To install development dependencies:

on Python 2.6:

pip install -r requirements_dev_2.6.txt

On later versions of Python:

pip install -r requirements_dev.txt

These will also install the runtime dependencies (i.e. requirements.txt, which just references the ‘install_requires’ field in setup.py) so you only ever need to install one requirements file.

To run tests:

make test

See the content of Makefile for a cheatsheet of other commonly used commands I use while working on rerun.

TODO

Print a timestamp at the top of the screen, so user can see when the command was last run.

Let user press a key to force a rerun.

Should this just be broken down into a command that waits for filesystem events and exits? Arguably calling “command” from the shell is many times easier than calling it from within “rerun”. Then ‘rerun’ could be written simply as a Bash function. Other bash functions could use the same pattern to iterate after waiting for a keypress, or a sleep (to poll.)

Known Problems

See issues at https://github.com/tartley/rerun/issues

Alternatives

PyPI package ‘watchdog’ is a cross-platform library for handling file-system events, which includes script ‘watchmedo’, which looks like a more serious and heavy-duty version of ‘Rerun’.

https://pypi-hypernode.com/pypi/watchdog

However, I can’t get watchmedo to work for my primary use case, which is re-running tests when files are saved by Vim. This is due to the way Vim writes to temporary files and then moves the temp file to overwrite data atomically. Vim thus guarantees that the user never loses data, but also fails to generate the correct FS events that watchmedo is looking for.

Thanks

Huge thanks to StackOverflow user ‘rici’, for explaining how to interpret shell aliases and functions in the given command using an interactive shell, and then how to retrieve the lost terminal once that shell terminates. http://stackoverflow.com/questions/25099895/from-python-start-a-shell-that-can-interpret-functions-and-aliases

Thanks to Bitbucket user sgourley for reporting, chasing up and even offering to fix issue #1, an egregious bug in ignoring directories, before I finally spotted the error. Thanks for the prods!

The idea came from the Bash command ‘watch’, and inspiration for this implementation came from an old blog post by Jeff Winkler, whos website http://jeffwinkler.net seems to have now died.

Changes

1.0.28

Add .cache (pytest) and ‘node_modules’ to the default ignored dirs.

1.0.27

Deleted version, I messed up the PyPI upload.

1.0.26

Prevent crash when a watched file gets deleted.

1.0.25

Documentation tweaks.

1.0.24

Last version broke rerun on Windows. Fixed now. (module ‘pwd’, which we started using to identify the user’s default shell, doesn’t exist on Windows.) So ‘-I’ or ‘–interactive’ won’t work on Windows, but probably the distinction between a regular shell and an interactive one is meaningless there anyway.

1.0.23

Running the given command using a non-interactive shell is now the default behaviour. Running an interactive shell for each command invocation requires sourcing .bashrc and the like, and hence is is slow, noisy on stdout, and error-prone. For example, virtualenvwrapper often fails to initialise properly and generates lots of output about it. An interactive shell can be requested when required (i.e. when the command to be run contains shell aliases or functions) with ‘–interactive’ or ‘-I’.

1.0.22

No longer crashes when a broken symlink exists under the cwd.

1.0.21

Run user’s given command in an interactive shell, so that shell aliases and functions are correctly interpreted.

1.0.20

(First release since migration to github) Fix to run under Python 3 again, specifically tested on 3.4.

1.0.19

Now expects commands to be a single arg (i.e. spaces etc should be escaped, or the whole command quoted) thus allowing rerun to work on composite commands, such as pipelines.

1.0.18

Now runs on Python 2.6, and is tested on Python 3.3.

Contact

Documentation & download:

http://pypi.python.org/pypi/rerun/

Souce code and issues:

https://github.com/tartley/rerun

Contact the author:

Jonathan Hartley, email: tartley at domain tartley.com, Twitter: @tartley.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

rerun-1.0.30.zip (19.2 kB view details)

Uploaded Source

File details

Details for the file rerun-1.0.30.zip.

File metadata

  • Download URL: rerun-1.0.30.zip
  • Upload date:
  • Size: 19.2 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: Python-urllib/2.7

File hashes

Hashes for rerun-1.0.30.zip
Algorithm Hash digest
SHA256 33bf86cb3d9dcdb51c6a6712b0cefcdebd7cdffce654d6bc9c8d7aae51e500e9
MD5 f5c56343b0172fa02bd1c837a1195139
BLAKE2b-256 95d9dd0334d2517b8ab0b3de0fff3e0370902c14057b47d2e15207a327a433a2

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page