hunter 3.0.5

Hunter is a flexible code tracing toolkit.

Hunter is a flexible code tracing toolkit, not for measuring coverage, but for debugging, logging, inspection and other nefarious purposes. It has a simple Python API, a convenient terminal API and a CLI tool to attach to processes.

• Free software: BSD 2-Clause License

Installation

pip install hunter

Documentation

https://python-hunter.readthedocs.io/

Overview

Basic use involves passing various filters to the trace option. An example:

import hunter
hunter.trace(module='posixpath', action=hunter.CallPrinter)

import os
os.path.join('a', 'b')

That would result in:

>>> os.path.join('a', 'b')
         /usr/lib/python3.6/posixpath.py:75    call      => join(a='a')
         /usr/lib/python3.6/posixpath.py:80    line         a = os.fspath(a)
         /usr/lib/python3.6/posixpath.py:81    line         sep = _get_sep(a)
         /usr/lib/python3.6/posixpath.py:41    call         => _get_sep(path='a')
         /usr/lib/python3.6/posixpath.py:42    line            if isinstance(path, bytes):
         /usr/lib/python3.6/posixpath.py:45    line            return '/'
         /usr/lib/python3.6/posixpath.py:45    return       <= _get_sep: '/'
         /usr/lib/python3.6/posixpath.py:82    line         path = a
         /usr/lib/python3.6/posixpath.py:83    line         try:
         /usr/lib/python3.6/posixpath.py:84    line         if not p:
         /usr/lib/python3.6/posixpath.py:86    line         for b in map(os.fspath, p):
         /usr/lib/python3.6/posixpath.py:87    line         if b.startswith(sep):
         /usr/lib/python3.6/posixpath.py:89    line         elif not path or path.endswith(sep):
         /usr/lib/python3.6/posixpath.py:92    line         path += sep + b
         /usr/lib/python3.6/posixpath.py:86    line         for b in map(os.fspath, p):
         /usr/lib/python3.6/posixpath.py:96    line         return path
         /usr/lib/python3.6/posixpath.py:96    return    <= join: 'a/b'
'a/b'

In a terminal it would look like:

Actions

Output format can be controlled with “actions”. There’s an alternative CodePrinter action that doesn’t handle nesting (it was the default action until Hunter 2.0).

If filters match then action will be run. Example:

import hunter
hunter.trace(module='posixpath', action=hunter.CodePrinter)

import os
os.path.join('a', 'b')

That would result in:

>>> os.path.join('a', 'b')
         /usr/lib/python3.6/posixpath.py:75    call      def join(a, *p):
         /usr/lib/python3.6/posixpath.py:80    line          a = os.fspath(a)
         /usr/lib/python3.6/posixpath.py:81    line          sep = _get_sep(a)
         /usr/lib/python3.6/posixpath.py:41    call      def _get_sep(path):
         /usr/lib/python3.6/posixpath.py:42    line          if isinstance(path, bytes):
         /usr/lib/python3.6/posixpath.py:45    line              return '/'
         /usr/lib/python3.6/posixpath.py:45    return            return '/'
                                               ...       return value: '/'
         /usr/lib/python3.6/posixpath.py:82    line          path = a
         /usr/lib/python3.6/posixpath.py:83    line          try:
         /usr/lib/python3.6/posixpath.py:84    line              if not p:
         /usr/lib/python3.6/posixpath.py:86    line              for b in map(os.fspath, p):
         /usr/lib/python3.6/posixpath.py:87    line                  if b.startswith(sep):
         /usr/lib/python3.6/posixpath.py:89    line                  elif not path or path.endswith(sep):
         /usr/lib/python3.6/posixpath.py:92    line                      path += sep + b
         /usr/lib/python3.6/posixpath.py:86    line              for b in map(os.fspath, p):
         /usr/lib/python3.6/posixpath.py:96    line          return path
         /usr/lib/python3.6/posixpath.py:96    return        return path
                                               ...       return value: 'a/b'
'a/b'

• or in a terminal:

---

Another useful action is the VarsPrinter:

import hunter
# note that this kind of invocation will also use the default `CallPrinter` action
hunter.trace(hunter.Q(module='posixpath', action=hunter.VarsPrinter('path')))

import os
os.path.join('a', 'b')

That would result in:

>>> os.path.join('a', 'b')
     /usr/lib/python3.6/posixpath.py:75    call      => join(a='a')
     /usr/lib/python3.6/posixpath.py:80    line         a = os.fspath(a)
     /usr/lib/python3.6/posixpath.py:81    line         sep = _get_sep(a)
     /usr/lib/python3.6/posixpath.py:41    call      [path => 'a']
     /usr/lib/python3.6/posixpath.py:41    call         => _get_sep(path='a')
     /usr/lib/python3.6/posixpath.py:42    line      [path => 'a']
     /usr/lib/python3.6/posixpath.py:42    line            if isinstance(path, bytes):
     /usr/lib/python3.6/posixpath.py:45    line      [path => 'a']
     /usr/lib/python3.6/posixpath.py:45    line            return '/'
     /usr/lib/python3.6/posixpath.py:45    return    [path => 'a']
     /usr/lib/python3.6/posixpath.py:45    return       <= _get_sep: '/'
     /usr/lib/python3.6/posixpath.py:82    line         path = a
     /usr/lib/python3.6/posixpath.py:83    line      [path => 'a']
     /usr/lib/python3.6/posixpath.py:83    line         try:
     /usr/lib/python3.6/posixpath.py:84    line      [path => 'a']
     /usr/lib/python3.6/posixpath.py:84    line         if not p:
     /usr/lib/python3.6/posixpath.py:86    line      [path => 'a']
     /usr/lib/python3.6/posixpath.py:86    line         for b in map(os.fspath, p):
     /usr/lib/python3.6/posixpath.py:87    line      [path => 'a']
     /usr/lib/python3.6/posixpath.py:87    line         if b.startswith(sep):
     /usr/lib/python3.6/posixpath.py:89    line      [path => 'a']
     /usr/lib/python3.6/posixpath.py:89    line         elif not path or path.endswith(sep):
     /usr/lib/python3.6/posixpath.py:92    line      [path => 'a']
     /usr/lib/python3.6/posixpath.py:92    line         path += sep + b
     /usr/lib/python3.6/posixpath.py:86    line      [path => 'a/b']
     /usr/lib/python3.6/posixpath.py:86    line         for b in map(os.fspath, p):
     /usr/lib/python3.6/posixpath.py:96    line      [path => 'a/b']
     /usr/lib/python3.6/posixpath.py:96    line         return path
     /usr/lib/python3.6/posixpath.py:96    return    [path => 'a/b']
     /usr/lib/python3.6/posixpath.py:96    return    <= join: 'a/b'
'a/b'

In a terminal it would look like:

---

You can give it a tree-like configuration where you can optionally configure specific actions for parts of the tree (like dumping variables or a pdb set_trace):

from hunter import trace, Q, Debugger
from pdb import Pdb

trace(
    # drop into a Pdb session if ``foo.bar()`` is called
    Q(module="foo", function="bar", kind="call", action=Debugger(klass=Pdb))
    |  # or
    Q(
        # show code that contains "mumbo.jumbo" on the current line
        lambda event: event.locals.get("mumbo") == "jumbo",
        # and it's not in Python's stdlib
        stdlib=False,
        # and it contains "mumbo" on the current line
        source__contains="mumbo"
    )
)

import foo
foo.func()

With a foo.py like this:

def bar():
    execution_will_get_stopped  # cause we get a Pdb session here

def func():
    mumbo = 1
    mumbo = "jumbo"
    print("not shown in trace")
    print(mumbo)
    mumbo = 2
    print(mumbo) # not shown in trace
    bar()

We get:

>>> foo.func()
not shown in trace
    /home/ionel/osp/python-hunter/foo.py:8     line          print(mumbo)
jumbo
    /home/ionel/osp/python-hunter/foo.py:9     line          mumbo = 2
2
    /home/ionel/osp/python-hunter/foo.py:1     call      def bar():
> /home/ionel/osp/python-hunter/foo.py(2)bar()
-> execution_will_get_stopped  # cause we get a Pdb session here
(Pdb)

In a terminal it would look like:

Tracing processes

In similar fashion to strace Hunter can trace other processes, eg:

hunter-trace --gdb -p 123

If you wanna play it safe (no messy GDB) then add this in your code:

from hunter import remote
remote.install()

Then you can do:

hunter-trace -p 123

See docs on the remote feature.

Note: Windows ain’t supported.

Environment variable activation

For your convenience environment variable activation is available. Just run your app like this:

PYTHONHUNTER="module='os.path'" python yourapp.py

On Windows you’d do something like:

set PYTHONHUNTER=module='os.path'
python yourapp.py

The activation works with a clever .pth file that checks for that env var presence and before your app runs does something like this:

from hunter import *
trace(<whatever-you-had-in-the-PYTHONHUNTER-env-var>)

Note that Hunter is activated even if the env var is empty, eg: PYTHONHUNTER="".

Environment variable configuration

Sometimes you always use the same options (like stdlib=False or force_colors=True). To save typing you can set something like this in your environment:

PYTHONHUNTERCONFIG="stdlib=False,force_colors=True"

This is the same as PYTHONHUNTER="stdlib=False,action=CallPrinter(force_colors=True)".

Notes:

• Setting PYTHONHUNTERCONFIG alone doesn’t activate hunter.

• All the options for the builtin actions are supported.

• Although using predicates is supported it can be problematic. Example of setup that won’t trace anything:

  PYTHONHUNTERCONFIG="Q(module_startswith='django')"
  PYTHONHUNTER="Q(module_startswith='celery')"

  which is the equivalent of:

  PYTHONHUNTER="Q(module_startswith='django'),Q(module_startswith='celery')"

  which is the equivalent of:

  PYTHONHUNTER="Q(module_startswith='django')&Q(module_startswith='celery')"

Filtering DSL

Hunter supports a flexible query DSL, see the introduction.

Development

To run the all tests run:

tox

FAQ

Why not Smiley?

There’s some obvious overlap with smiley but there are few fundamental differences:

• Complexity. Smiley is simply over-engineered:

  • It uses IPC and a SQL database.

  • It has a webserver. Lots of dependencies.

  • It uses threads. Side-effects and subtle bugs are introduced in your code.

  • It records everything. Tries to dump any variable. Often fails and stops working.

  Why do you need all that just to debug some stuff in a terminal? Simply put, it’s a nice idea but the design choices work against you when you’re already neck-deep into debugging your own code. In my experience Smiley has been very buggy and unreliable. Your mileage may vary of course.

• Tracing long running code. This will make Smiley record lots of data, making it unusable.

  Now because Smiley records everything, you’d think it’s better suited for short programs. But alas, if your program runs quickly then it’s pointless to record the execution. You can just run it again.

  It seems there’s only one situation where it’s reasonable to use Smiley: tracing io-bound apps remotely. Those apps don’t execute lots of code, they just wait on network so Smiley’s storage won’t blow out of proportion and tracing overhead might be acceptable.

• Use-cases. It seems to me Smiley’s purpose is not really debugging code, but more of a “non interactive monitoring” tool.

In contrast, Hunter is very simple:

• Few dependencies.

• Low overhead (tracing/filtering code has an optional Cython extension).

• No storage. This simplifies lots of things.

  The only cost is that you might need to run the code multiple times to get the filtering/actions right. This means Hunter is not really suited for “post-mortem” debugging. If you can’t reproduce the problem anymore then Hunter won’t be of much help.

Why not pytrace?

Pytrace is another tracer tool. It seems quite similar to Smiley - it uses a sqlite database for the events, threads and IPC.

TODO: Expand this.

Why (not) coverage?

For purposes of debugging coverage is a great tool but only as far as “debugging by looking at what code is (not) run”. Checking branch coverage is good but it will only get you as far.

From the other perspective, you’d be wondering if you could use Hunter to measure coverage-like things. You could do it but for that purpose Hunter is very “rough”: it has no builtin storage. You’d have to implement your own storage. You can do it but it wouldn’t give you any advantage over making your own tracer if you don’t need to “pre-filter” whatever you’re recording.

In other words, filtering events is the main selling point of Hunter - it’s fast (cython implementation) and the query API is flexible enough.

Changelog

3.0.5 (2019-12-06)

• Really fixed safe_repr so it doesn’t cause side-effects (now isinstance/issubclass are avoided - they can cause side-effects in code that abuses descriptors in special attributes/methods).

3.0.4 (2019-10-26)

• Really fixed stream setup in actions (using force_colors without any stream was broken). See: hunter.actions.ColorStreamAction.

• Fixed __repr__ for the hunter.predicates.From predicate to include watermark.

• Added binary wheels for Python 3.8.

3.0.3 (2019-10-13)

• Fixed safe_repr on pypy so it’s safer on method objects. See: hunter.actions.ColorStreamAction.

3.0.2 (2019-10-10)

• Fixed setting stream from PYTHONHUNTERCONFIG environment variable. See: hunter.actions.ColorStreamAction.

• Fixed a couple minor documentation issues.

3.0.1 (2019-06-17)

• Fixed issue with coloring missing source message (coloring leaked into next line).

3.0.0 (2019-06-17)

• The package now uses setuptools-scm for development builds (available at https://test.pypi.org/project/hunter/). As a consequence installing the sdist will download setuptools-scm.

• Recompiled cython modules with latest Cython. Hunter can be installed without any Cython, as before.

• Refactored some of the cython modules to have more typing information and not use deprecated property syntax.

• Replaced unsafe_repr option with repr_func. Now you can use your custom repr function in the builtin actions. BACKWARDS INCOMPATIBLE

• Fixed buggy filename handling when using Hunter in ipython/jupyter. Source code should be properly displayed now.

• Removed globals option from VarsPrinter action. Globals are now always looked up. BACKWARDS INCOMPATIBLE

• Added support for locals in VarsPrinter action. Now you can do VarsPrinter('len(foobar)').

• Always pass module_globals dict to linecache methods. Source code from PEP-302 loaders is now printed properly. Contributed by Mikhail Borisov in #65.

• Various code cleanup, style and docstring fixing.

• Added hunter.From helper to allow passing in filters directly as keyword arguments.

• Added hunter.event.Event.detach for storing events without leaks or side-effects (due to prolonged references to Frame objects, local or global variables).

• Refactored the internals of actions for easier subclassing.

  Added the hunter.actions.ColorStreamAction.filename_prefix, hunter.actions.ColorStreamAction.output, hunter.actions.ColorStreamAction.pid_prefix, hunter.actions.ColorStreamAction.thread_prefix, hunter.actions.ColorStreamAction.try_repr and hunter.actions.ColorStreamAction.try_source methods to the hunter.actions.ColorStreamAction baseclass.

• Added hunter.actions.VarsSnooper - a PySnooper-inspired variant of hunter.actions.VarsPrinter. It will record and show variable changes, with the risk of leaking or using too much memory of course :)

• Fixed tracers to log error and automatically stop if there’s an internal failure. Previously error may have been silently dropped in some situations.

2.2.1 (2019-01-19)

• Fixed a link in changelog.

• Fixed some issues in the Travis configuration.

2.2.0 (2019-01-19)

• Added hunter.predicates.From predicate for tracing from a specific point. It stop after returning back to the same call depth with a configurable offset.

• Fixed PYTHONHUNTERCONFIG not working in some situations (config values were resolved at the wrong time).

• Made tests in CI test the wheel that will eventually be published to PyPI (tox-wheel).

• Made event.stdlib more reliable: pkg_resources is considered part of stdlib and few more paths will be considered as stdlib.

• Dumbed down the get_peercred check that is done when attaching with hunter-trace CLI (via hunter.remote.install()). It will be slightly insecure but will work on OSX.

• Added OSX in the Travis test grid.

2.1.0 (2018-11-17)

• Made threading_support on by default but output automatic (also, now 1 or 0 allowed).

• Added pid_alignment and force_pid action options to show a pid prefix.

• Fixed some bugs around __eq__ in various classes.

• Dropped Python 3.3 support.

• Dropped dependency on fields.

• Actions now repr using a simplified implementation that tries to avoid calling __repr__ on user classes in order to avoid creating side-effects while tracing.

• Added support for the PYTHONHUNTERCONFIG environment variable (stores defaults and doesn’t activate hunter).

2.0.2 (2017-11-24)

• Fixed indentation in hunter.actions.CallPrinter action (shouldn’t deindent on exception).

• Fixed option filtering in Cython Query implementation (filtering on tracer was allowed by mistake).

• Various fixes to docstrings and docs.

2.0.1 (2017-09-09)

• Now Py_AddPendingCall is used instead of acquiring the GIL (when using GDB).

2.0.0 (2017-09-02)

• Added the hunter.event.Event.count and hunter.event.Event.calls` attributes.

• Added the lt/lte/gt/gte lookups.

• Added convenience aliases for startswith (sw), endswith (ew), contains (has) and regex (rx).

• Added a convenience hunter.wrap decorator to start tracing around a function.

• Added support for remote tracing (with two backends: manhole and GDB) via the hunter-trace bin. Note: Windows is NOT SUPPORTED.

• Changed the default action to hunter.actions.CallPrinter. You’ll need to use action=CodePrinter if you want the old output.

1.4.1 (2016-09-24)

• Fix support for getting sources for Cython module (it was broken on Windows and Python3.5+).

1.4.0 (2016-09-24)

• Added support for tracing Cython modules (#30). A # cython: linetrace=True stanza or equivalent is required in Cython modules for this to work.

1.3.0 (2016-04-14)

• Added hunter.event.Event.thread.

• Added hunter.event.Event.threadid and hunter.event.Event.threadname (available for filtering with hunter.Q).

• Added hunter.event.Event.threading_support argument to hunter.trace. It makes new threads be traced and changes action output to include thread name.

• Added support for using pdb++ in the hunter.actions.Debugger action.

• Added support for using manhole via a new hunter.actions.Manhole action.

• Made the hunter.event.Event.handler a public but readonly property.

1.2.2 (2016-01-28)

• Fix broken import. Require fields>=4.0.

• Simplify a string check in Cython code.

1.2.1 (2016-01-27)

• Fix “KeyError: ‘normal’” bug in hunter.actions.CallPrinter. Create the NO_COLORS dict from the COLOR dicts. Some keys were missing.

1.2.0 (2016-01-24)

• Fixed printouts of objects that return very large string in __repr__(). Trimmed to 512. Configurable in actions with the repr_limit option.

• Improved validation of hunter.actions.VarsPrinter’s initializer.

• Added a hunter.actions.CallPrinter action.

1.1.0 (2016-01-21)

• Implemented a destructor (__dealloc__) for the Cython tracer.

• Improved the restoring of the previous tracer in the Cython tracer (use PyEval_SetTrace) directly.

• Removed tracer as an allowed filtering argument in hunter.Query.

• Add basic validation (must be callable) for positional arguments and actions passed into hunter.Q. Closes #23.

• Fixed stdlib checks (wasn’t very reliable). Closes #24.

1.0.2 (2016-01-05)

• Fixed missing import in setup.py.

1.0.1 (2015-12-24)

• Fix a compile issue with the MSVC compiler (seems it don’t like the inline option on the fast_When_call).

1.0.0 (2015-12-24)

• Implemented fast tracer and query objects in Cython. MAY BE BACKWARDS INCOMPATIBLE

  To force using the old pure-python implementation set the PUREPYTHONHUNTER environment variable to non-empty value.

• Added filtering operators: contains, startswith, endswith and in. Examples:

  • Q(module_startswith='foo' will match events from foo, foo.bar and foobar.

  • Q(module_startswith=['foo', 'bar'] will match events from foo, foo.bar, foobar, bar, bar.foo and baroo .

  • Q(module_endswith='bar' will match events from foo.bar and foobar.

  • Q(module_contains='ip' will match events from lipsum.

  • Q(module_in=['foo', 'bar'] will match events from foo and bar.

  • Q(module_regex=r"(re|sre.*)\b") will match events from ``re, re.foobar, srefoobar but not from repr.

• Removed the merge option. Now when you call hunter.trace(...) multiple times only the last one is active. BACKWARDS INCOMPATIBLE

• Remove the previous_tracer handling. Now when you call hunter.trace(...) the previous tracer (whatever was in sys.gettrace()) is disabled and restored when hunter.stop() is called. BACKWARDS INCOMPATIBLE

• Fixed CodePrinter to show module name if it fails to get any sources.

0.6.0 (2015-10-10)

• Added a clear_env_var option on the tracer (disables tracing in subprocess).

• Added force_colors option on hunter.actions.VarsPrinter and hunter.actions.CodePrinter.

• Allowed setting the stream to a file name (option on hunter.actions.VarsPrinter and hunter.actions.CodePrinter).

• Bumped up the filename alignment to 40 cols.

• If not merging then self is not kept as a previous tracer anymore. Closes #16.

• Fixed handling in VarsPrinter: properly print eval errors and don’t try to show anything if there’s an AttributeError. Closes #18.

• Added a stdlib boolean flag (for filtering purposes). Closes #15.

• Fixed broken frames that have “None” for filename or module (so they can still be treated as strings).

• Corrected output files in the install_lib command so that pip can uninstall the pth file. This only works when it’s installed with pip (sadly, setup.py install/develop and pip install -e will still leave pth garbage on pip uninstall hunter).

0.5.1 (2015-04-15)

• Fixed hunter.event.Event.globals to actually be the dict of global vars (it was just the locals).

0.5.0 (2015-04-06)

• Fixed hunter.And and hunter.Or “single argument unwrapping”.

• Implemented predicate compression. Example: Or(Or(a, b), c) is converted to Or(a, b, c).

• Renamed hunter.event.Event.source to hunter.event.Event.fullsource.

• Added hunter.event.Event.source that doesn’t do any fancy sourcecode tokenization.

• Fixed hunter.event.Event.fullsource return value for situations where the tokenizer would fail.

• Made the print function available in the PYTHONHUNTER env var payload.

• Added a __repr__ for hunter.event.Event.

0.4.0 (2015-03-29)

• Disabled colors for Jython. Contributed by Claudiu Popa in #12.

• Test suite fixes for Windows. Contributed by Claudiu Popa in #11.

• Added an introduction section in the docs.

• Implemented a prettier fallback for when no sources are available for that frame.

• Implemented fixups in cases where you use action classes as a predicates.

0.3.1 (2015-03-29)

• Forgot to merge some commits …

0.3.0 (2015-03-29)

• Added handling for internal repr failures.

• Fixed issues with displaying code that has non-ascii characters.

• Implemented better display for call frames so that when a function has decorators the function definition is shown (instead of just the first decorator). See: #8.

0.2.1 (2015-03-28)

• Added missing color entry for exception events.

• Added hunter.event.Event.line property. It returns the source code for the line being run.

0.2.0 (2015-03-27)

• Added color support (and colorama as dependency).

• Added support for expressions in hunter.actions.VarsPrinter.

• Breaking changes:

  • Renamed F to hunter.Q. And hunter.Q is now just a convenience wrapper for hunter.predicates.Query.

  • Renamed the PYTHON_HUNTER env variable to PYTHONHUNTER.

  • Changed hunter.predicates.When to take positional arguments.

  • Changed output to show 2 path components (still not configurable).

  • Changed hunter.actions.VarsPrinter to take positional arguments for the names.

• Improved error reporting for env variable activation (PYTHONHUNTER).

• Fixed env var activator (the .pth file) installation with setup.py install (the “egg installs”) and setup.py develop/pip install -e (the “egg links”).

0.1.0 (2015-03-22)

• First release on PyPI.