Pieshell is a Python shell environment that combines the expressiveness of shell pipelines with the power of python iterators.
It can be used in two major ways:
- As an interactive shell replacing e.g. bash
- As an ordinary python module replacing e.g. subprocess.Popen
- Installation
- As a shell
- As a python module
- Configuration
- Builtins
- External tools
- Tips and tricks
- Unit tests
- Copyright
$ pip install pieshell
To start pieshell in interactive mode, just run the command pieshell:
$ pieshell
The interactive pieshell environment supports all normal python syntax.
140:/home/redhog/Projects/beta/pieshell >>> print 3+4
7
In addition, you can run programs just like in any shell by writing their names
140:/home/redhog/Projects/beta/pieshell >>> ls
build deps dist LICENSE.txt pieshell pieshell.egg-info README.md
setup.py
Parameters to programs however have to be given as proper python strings within parenthesis, like a python function call
140:/home/redhog/Projects/beta/pieshell >>> ls("-a")
. .. build deps dist .git .gitignore LICENSE.txt pieshell
pieshell.egg-info .#README.md README.md setup.py
Piping the standard output of one command to the standard input of another works just like in bash
140:/home/redhog/Projects/beta/pieshell >>> ls("-a") | grep("-e", ".py")
setup.py
Changing directory is done using the command cd:
140:/home/redhog/Projects/beta/pieshell >>> cd("..")
140:/home/redhog/Projects/beta >>>
To execute commands that require a path, for example ones in the current directory, or commands with a dot in their names
140:/home/redhog/Projects/beta/pieshell >>> _("./setup.py", "--help")
Common commands: (see '--help-commands' for more)
...
The underscore represents the virtual root command that has no parameters, not even a command name. In general, there are two equivalent syntaxes for parameters: as function parameter strings, and as attribute names. The two syntaxes can be mixed freely. All of the following are equivalent:
_("foo", "bar", "fie")
_.foo("bar", "fie")
_.foo.bar.fie()
foo.bar.fie()
foo.bar.fie
Example usage:
git.diff("-U")
In addition to these two generic syntaxes, there are two more specialized syntaxes for options:
The function call syntax also supports named parameters, which are converted into "--name=value" pairs. Note that the order can not be guaranteed as named parameters are sent around as dictionaries inside python:
git.diff(unified=4)
Short options, like -U above do not actually need quotes, and can be
specified inside or outside the function call syntax:
git.diff(-U, -w)
git.diff-U-w
Notes for programs with wierd parameter syntax, like find: find
does not use the standard double minus (--) before long options, and
takes the option value as a separate argument, rather than separating
the name and value with = like most programs. The above special
syntax for short options actually cover this use case too:
find(".", -name, "*.py")
Standard out and standard in of a pipeline can be redirected to a file by piping to or from a string (the filename). As a special case, None is a short hand for "/dev/null"
140:/home/redhog/Projects/beta/pieshell >>> ls | "foo"
140:/home/redhog/Projects/beta/pieshell >>> "foo" | cat
bar
build
deps
dist
foo
LICENSE.txt
pieshell
pieshell.egg-info
README.md
setup.py
140:/home/redhog/Projects/beta/pieshell >>> ls | None
Redirects can also be made with a more explicit syntax that allows redirecting other file descriptors than stdin and stdout:
139:/home/redhog/Projects/beta/pieshell >>> cat |
Redirect("stdin", "foo") | Redirect("stdout", "bar")
The constructor for redirect takes the following arguments:
Redirect(fd, source, flag=None, mode=0777)
fd can be either an int, or one of "stdin", "stdout" and "stderr".
source is either a string filename, an int file descriptor or one of
the special values PIPE, TMP and STRING. PIPE create a new
pipe which you are required to connect up manually. TMP generates a
temporary named file. The filename is available in
RunningPipeline.processes[].output_files[filedescriptornr].path.
STRING generates a temporary named file, but reads and deletes it
when the pipeline exists, putting its content in
RunningPipeline.processes[].output_content[filedescriptornr].
flag and mode have the same semantics as for os.open(). Flags do
not have to be given for stdin, stdout and stderr / fd 0, 1
and 2 and defaults to os.O_RDONLY or os.O_RDONLY | os.O_CREAT.
Shell commands are first class python objects, and their input and output can be interacted with easily from python in the form of iterators. Iterating over a shell command iterates over the lines of its standard out
140:/home/redhog/Projects/beta/pieshell >>> list(ls("-a"))
['.', '..', 'build', 'deps', 'dist', '.git', '.gitignore',
'LICENSE.txt', 'pieshell', 'pieshell.egg-info', '.#README.md',
'README.md', 'setup.py']
140:/home/redhog/Projects/beta/pieshell >>> for x in ls("-a"):
... if x.endswith('.py'):
... print x
...
setup.py
Piping an iterator into a shell command, sends its items as lines to the standard in of the shell command
140:/home/redhog/Projects/beta/pieshell >>> list(["foo", "bar.py", "fie.py"] |
grep("-e", ".py"))
['bar.py', 'fie.py']
140:/home/redhog/Projects/beta/pieshell >>> def foo():
... yield "hello"
... yield "world"
140:/home/redhog/Projects/beta/pieshell >>> foo() | cat
hello
world
In addtion, iterators and pipelines may be used as arguments to commands and will be seen by the command as a filename, which when opened and read from will produce the output of that iterator as lines, or the standard output of the pipeline.
140:/home/redhog/Projects/beta/pieshell >>> list(cat(["foo", "bar"]))
['foo', 'bar']
140:/home/redhog/Projects/beta/pieshell >>> list(cat(["foo", "bar"] | cat))
['foo', 'bar']
Environment variables are available directly in the shell as variables, together with any local python variables. In addition, they are available in the dictionary exports.
140:/home/redhog/Projects/beta/pieshell >>> LANG
'en_US.UTF-8'
Assigning to the name of an already exported environment variable updates the value of that variable.
140:/home/redhog/Projects/beta/pieshell >>> LANG = "sv_SE.UTF-8"
140:/home/redhog/Projects/beta/pieshell >>> exports["LANG"]
'sv_SE.UTF-8'
Assigning to a variable name not already used as an environment variable creates a local python variable.
140:/home/redhog/Projects/beta/pieshell >>> foo = "hello"
140:/home/redhog/Projects/beta/pieshell >>> "foo" in exports
False
140:/home/redhog/Projects/beta/pieshell >>> foo
'hello'
To export a new variable, you have to assign it in the exports dictionary.
140:/home/redhog/Projects/beta/pieshell >>> exports["bar"] = "world"
140:/home/redhog/Projects/beta/pieshell >>> bar
'world'
All parameter strings in commands are subject to expansion unless wrapped in a call to R(), e.g. R("my * string * here").
-
"~" and "~username" are expanded using os.path.expanduser()
-
Variable expansion is done using the python % operator on python variables as well as environment variables.
-
Pattern matching is done using glob.glob()
A running pipeline is represented by a RunningPipeline instance. This object is returned by the Pipeline.run() and Pipeline.run_interactive() methods. In interactive shell mode the RunningPipeline instance for the last executed pipeline is available in the last_pipeline variable.
A RunningPipeline instance can be used to extract events and statuses of the processes involved in the pipeline:
-
RunningPipeline.processes is a list of RunningItem instances, each representing an external process or a python function.
-
RunningPipeline.failed_processes is a list of RunningItem instances for those processes in the pipeline that have failed (returned a non-zero exit status).
-
RunningPipeline.pipeline is a (deep) copy of the original pipeline object, with additional run status added, e.g. links to processes, exit status etc.
-
RunningPipeline.wait() waits for all processes in the pipeline to terminate.
As a convenience, RunningItem:s can be retrieved by the name of their corresponding binary as properties of the RunningPipeline. E.g.
p = ls | grep(-e, ".py")
p.grep
A RunningItem instance represents an external process or a python function:
-
RunningItem.cmd points to the part of the RunningPipeline.pipeline structure that gave rise to this process.
-
RunningItem.is_running is True if the process is still running.
-
RunningItem.is_failed is True if the process has failed somehow (process with non-zero exit status, function threw an exception).
-
RunningItem.output_content contains a dictionary of the output of any STRING redirection for the process with the file descriptors as keys.
-
RunningProcess.iohandler.last_event contains a dictionary of the members of the last event from the process. The members have the same names and meaning as the members of the signalfd_siginfo struct, see "man signalfd" for details.
This functionality is only available if psutil is installed.
The class PstreeProcess represents any process not directly started
by Pieshell.
RunningPipeline, RunningItem, and PstreeProcess all exposes a
unified interface where child processes, parent processes, group
leaders and session leaders are available as properties. Child
porcesses uses the names of their binaries as property names. Parent
processes, session and group leaders use PARENT, SESS and GROUP
respectively.
RunningItem.details, and PstreeProcess.details contains a
psutil.Process() instance, and most of the members and methods of
that instance are directly available on the RunningItem or
PstreeProcess object.
To access a PstreeProcess object for the current shell and for the
init process (pid 1), there are two global variables INIT and
CURRENT.
The currently logged in users and their login session processes can be
accessed using the global USERS variable with a similar semantic for
properties.
All these objects support tab completion to let you easily naviage the tree of running porcesses on your system.
redhog@glittertind:~/pieshell[master] >>> USERS.redhog.localhost.mate_panel.mate_terminal.bash.bin_bash.bash.pid_10171.python3_9
_('/home/redhog/pieshell/env/bin/python3.9', '/home/redhog/pieshell/env/bin/pieshell') as 26341
redhog@glittertind:~/pieshell[master] >>> CURRENT
_('/home/redhog/pieshell/env/bin/python3.9', '/home/redhog/pieshell/env/bin/pieshell') as 26341
A pipeline can be started in the background by appending &True, or
&None to do the same and also redirect stdout to /dev/null.
last_pipeline can be used to access the backgrounded pipeline.
A running pipeline can be stopped by hitting CTRL-Z. A stopped
pipeline can be restarted in the background with any of
bg
bg(last_pipeline)
last_pipeline.restart()
or in the foreground with
fg fg(last_pipeline) last_pipeline.wait()
When a pipeline fails, e.g. by one of the involved processes exiting with a non-zero status, RunningPipeline.wait() and Pipeline.run_interactive() will throw a PipelineFailed exception after all processes have exited.
- PipelineFailed.pipeline holds a reference to the RunningPipeline instance that generated the exception.
If a pipeline is interrupted with CTRL-C, a PipelineInterrupted is raised.
- PipelineInterrupted.pipeline holds a reference to the RunningPipeline instance.
If you want to catch errors in a script, you can use normal Python exception handling:
try:
except PipelineFailed as e:
e.pipeline.failed_processes[0].pipeline
Bash provides the command source to run the content of a bash script inside the current shell,
effectively letting an external script update the environment variables of the running shell.
This functionality is often used for setting up local development environments, like virtualenv.
Pieshell provides a builtin to emulate this functionality, with bash scripts:
>>> bashsource("myscript.sh")
will run myscript.sh in a bash shell followed by declare -x. It parses the output of
declare -x and updates exports accordingly. As a special case
>>> bashsource()
will do the same, but without running any script first, esentially
just using whatever variables are set up by your .bashrc or
.profile.
All constructs described above to use iterators, can equally well be used as / with async iterators. Pipelines can be awaited, which runs them and waits for them to exit. RunningPipelines can be awaited, which just waits for them to exit.
>>> from pieshell import *
All functionality available in the interactive shell is available when using pieshell as an ordinary python module. However, a slighly more cumbersome syntax is required.
In particular, shell commands can not be run just by writing their names. Instead, they have to be accessed as members of the "env" object:
>>> list(env.ls("-a") | env.grep("-e", "io"))
["iterio.py", "iterio.pyc"]
Commands are also not run with standard out to the screen when simply printed using the repr() function but must instead be used as iterators as is done above using the list() function.
A pipeline can be run in the background (with input/output to the screen) using
>>>> status = ~(env.ls("-a") | env.grep("-e", "io"))
or in the forground (waiting until it exists)
>>>> status = +(env.ls("-a") | env.grep("-e", "io"))
The env object holds the current working directory, which can be changed with
>>> env.cd("..")
You can also create multiple environments and use them siumultaneously, even within the same pipeline
>>> env2 = env()
>>> env2.cd("somedir")
Environment variables are available as a dictionary in env._exports.
Variable expansion is only done on environment variables, as there is no way for pieshell to find out about the right scope to do variable lookups in in any given situation.
In addition to being able to use pieshell code in ordinary python modules using this slightly more verbose syntax, pieshell supports importing modules named modulename.pysh rather than modulename.py. Pysh modules support the full syntax of the interactive pieshell console. Pysh modules can be imported using the standard import syntax as soon as pieshell itself has been imported, and from the interactive pieshell.
Be aware that you won't get a NameError for misspelled variable
names in pyshmodules, but instead a
pieshell.pipeline.running.PipelineFailed error, or if not used in a
way that will cause it to run, just a pipeline object stored somewhere
it shouldn't.
Sometimes you don't want to extract a bunch of functions to a separate
pysh module, but you'd still like to be able to use the simpler syntax
in them. That's still possible usig the @pyshfunction decorator:
>>> @pieshell.env.pyshfunction
>>> def foo():
... return list(echo("hello"))
You can use any environment for this, not only the default one like in the example above.
By giving an instance of pieshell.environ.EnvScope with an environment
set to interactive as globals argument to either eval or exec, the
full syntax of Pieshell can be used in the string to evaluate/execute.
>>> scope = pieshell.environ.EnvScope(env=pieshell.env(interactive=True))
>>> eval("ls()", scope)
build deps dist LICENSE.txt pieshell pieshell.egg-info README.md
setup.py
When running pieshell in interactive mode it executes
~/.config/pieshell at startup if it exists. This file can be used to
configure the interactive environment the same way ~/.bashrc can be
used to configure the bash shell. For example it can be used to load
python modules, execute shell pipelines or set environment variables.
An example config file is supplied in
pieshell/resources/default_config.pysh.
The prompt that Pieshell uses is generated by the method
pieshell.Environment.__str__. This method can be overridden by
assigning a new function to it:
def prompt(self):
return "Pieshell> "
Be careful if you run pipelines or other code inside the prompt that might call repr() as that might lead to an infinite loop of prompt printing. To solve this, use a thread local variable as a flag to catch and break any recursion.
While pieshell lets you pipe to and from ordinary python functions, they don't offer the same syntax and tab-completion as external commands (e.g. 'myfunction.arg1.arg2(name=value)'), they can't modify the environment or do fancy redirects. Builtin commands provide all of this, at the cost of a slightly clumsier syntax:
class MyMagicBuiltin(pieshell.Builtin):
"""More magic to the people
"""
name = "magic"
def _run(self, redirects, sess, indentation = ""):
# redirects is an instance of pieshell.Redirects
#
# sess is an opaque data structure that must be passed to
# any call to _run() you do yourself from this method (or
# any function it calls).
#
# indentation is a string containing only whitespace, to
# be prepended to any debug printing lines you print.
#
# Returns a list of instances of some pieshell.RunningItem
# subclass
self._cmd = self._env.find(
".", "-name", "%s.txt" % self._arg[1]) | self._env.tac
return self._cmd._run(redirects, sess, indentation)
# Optional for tab completion
def __dir__(self):
return ["light", "dark"]
Builtins need to be registered to be available in the shell. This can be done either from python with
pipeline.BuiltinRegistry.register(MyMagicBuiltin)
or from setup.py in any package:
setup(
entry_points={
'pieshell.builtin': [
"magic = mypackage:MyMagicBuiltin",
]
}
)
A short list of tools that might be usefull together with this project:
- psutil - python api for getting ps / top style information about a process
- ReRedirect - redirect io for an already running process
- Reptyr - move a running process to a new controlling terminal
- Deptyr - forward output for a new process to another controlling terminal
The pandas reader for a table of fixed-width formatted lines does a pretty good job at parsing the output from several standard command line tools such as ls(-l):
>>> pd.read_fwf(io.StringIO(str(ls(-l))))
total 328 Unnamed: 1 Unnamed: 2 Unnamed: 3 Unnamed: 4 Unnamed: 5 Unnamed: 6 Unnamed: 7 Unnamed: 8
0 drwxr-xr-x 6 redhog redhog 4096 okt. 16 21:45 build
1 -rw-rw-r-- 1 redhog redhog 25 jan. 11 2021 _config.yml
2 drwxrwxr-x 2 redhog redhog 4096 okt. 25 21:30 dist
3
>>> pd.read_fwf(io.StringIO(str(docker.images)))
REPOSITORY TAG IMAGE ID CREATED SIZE
0 python 3 a8405b7e74cf 7 months ago 921MB
1 ubuntu 22.04 08d22c0ceb15 7 months ago 77.8MB
2 ubuntu latest d13c942271d6 21 months ago 72.8MB
This pattern has been provided as convenience method on pipelines:
>>> docker.images.to_dataframe()
The package libsixel-bin lets you convert images to a format that can be directly displayed inside some terminal emulators. This can be easily combined with matplotlib to display data Jupyter Notebook inline style, but in your terminal:
>>> import matplotlib.pyplot as plt, io
>>> plt.scatter([1, 2, 3], [3, 2, 4])
>>> buf = io.BytesIO(); plt.savefig(buf, format = 'png')
>>> [buf.getvalue()] | img2sixel
<Image displayed here in supporting terminals>
To run the unit- and integration tests
pip install nose2
nose2 -s tests
Pieshell copyright 2016 Egil Möller [email protected]
Pieshell is free software: you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU Lesser General Public License along with this program. If not, see http://www.gnu.org/licenses/.