landscapeio / prospector Goto Github PK

If a directory beginning with '.' is on your current path, then prospector will fail to generate any messages.

at https://landscape.io/github/fsr-itse/EvaP/95/modules/evap/evaluation/meta.py#L19 it says

 def __new__(metacls, classname, bases, classDict):
    Metaclass class method __new__ should have 'mcs' as first argument
    first argument of a classmethod should be named 'cls'

so... yeah.

I've installed by cloning from git and then ran python setup.py install

Traceback (most recent call last):
  File "/usr/local/bin/prospector", line 9, in <module>
    load_entry_point('prospector==0.5.1', 'console_scripts', 'prospector')()
  File "/usr/local/lib/python2.7/dist-packages/prospector-0.5.1-py2.7.egg/prospector/run.py", line 186, in main
    summary, messages = prospector.execute()
  File "/usr/local/lib/python2.7/dist-packages/prospector-0.5.1-py2.7.egg/prospector/run.py", line 131, in execute
    tool.prepare(self.path, self.ignores, self.config, self.adaptors)
  File "/usr/local/lib/python2.7/dist-packages/prospector-0.5.1-py2.7.egg/prospector/tools/pylint/__init__.py", line 83, in prepare
    adaptor.adapt_pylint(linter)
  File "/usr/local/lib/python2.7/dist-packages/prospector-0.5.1-py2.7.egg/prospector/adaptor/common.py", line 8, in adapt_pylint
    linter.load_plugin_modules(['pylint_common'])
  File "/usr/local/lib/python2.7/dist-packages/pylint-1.1.0-py2.7.egg/pylint/lint.py", line 357, in load_plugin_modules
    module = load_module_from_name(modname)
  File "/usr/local/lib/python2.7/dist-packages/logilab_common-0.61.0-py2.7.egg/logilab/common/modutils.py", line 125, in load_module_from_name
    return load_module_from_modpath(dotted_name.split('.'), path, use_sys)
  File "/usr/local/lib/python2.7/dist-packages/logilab_common-0.61.0-py2.7.egg/logilab/common/modutils.py", line 167, in load_module_from_modpath
    mp_file, mp_filename, mp_desc = find_module(part, path)
ImportError: No module named pylint_common

So I tried:

sjuul@development01:~/workspace/oversight$ sudo pip install -U pylint-common
Requirement already up-to-date: pylint-common in /usr/local/lib/python2.7/dist-packages/pylint_common-0.1-py2.7.egg
Requirement already up-to-date: pylint>=1.0 in /usr/local/lib/python2.7/dist-packages/pylint-1.1.0-py2.7.egg (from pylint-common)
Requirement already up-to-date: astroid>=1.0 in /usr/local/lib/python2.7/dist-packages/astroid-1.0.1-py2.7.egg (from pylint-common)
Requirement already up-to-date: pylint-plugin-utils>=0.1 in /usr/local/lib/python2.7/dist-packages/pylint_plugin_utils-0.1.1-py2.7.egg (from pylint-common)
Requirement already up-to-date: logilab-common>=0.53.0 in /usr/local/lib/python2.7/dist-packages/logilab_common-0.61.0-py2.7.egg (from pylint>=1.0->pylint-common)

But that didn't help

On projects which use tox (it complements Travis-CI in the same way an offline copy of Prospector complements Landscape.io), the user is forced to choose between not inspecting files like setup.py in the root of the repository (which differs from my understanding of how Landscape.io does things) and a very awkward hack-around involving temporarily moving .tox outside the project repository.

(Since letting it inspect .tox takes forever and produces a flood of spurious errors from the project's dependencies in the virtualenvs)

There needs to be some kind of mechanism for ignoring folders and .tox should be included in the default setting.

The Python 3 Wall of Superpowers features the most downloaded 210 modules of PyPI. At the end of the list, there are still very popular packages like bottle, Zope or Pylons. Prospector should support all of them to help with many use cases.

This is related to #56. It would be easier to wrap prospector in other tools (such as a git commit hook) if it could accept more than one path at a time.

I'm interested in implementing this myself and submitting it as a pull request, but I wouldn't mind if someone else with more free time does it instead.

The python-skeletons project has a lot of goals in common with prospector's attempts to adapt static analyis tools to particular frameworks. Ideally, pylint-django and pylint-celery could be subsumed into pylint-skeletons, and prospector would include pylint-skeletons.

This issue is a rather vague "investigate this" reminder rather than anything specific.

Prospector should use existing pylint config if available

diffcov automatically finds the diff lines that needs tests.
It also finds lint errors on diffs which is very useful for PRs.

(moving from landscapeio/landscape-issues#67)

Currently prospector looks for files with a .py extension, but this could be a bit smarter. For example, first use mimetype to look for text files, then inspect those for a python hashbang or other indications of python-ness.

I get TypeError: unorderable types: str() > NoneType() upon running prospector on requirements.txt. This is python 3.4 with prospector 0.7.

My requirements.txt: https://gist.github.com/374113a40502b04c5112

Did some debug and found out that it tries to compare class pylint.checkers.base.BasicChecker with instance of site-packages.pylint.checkers.base.BasicChecker

How I could fix it?

just noticed that the linter appears to not display multiple instances of an error on the same page.

if i have

import json, pprint, 

a = 1
b = 1
print a + b

I will get 2 errors, one E401, and one W0611 for pprint.
However I should have 3, E401, W0611 for json, and W0611 for pprint.

Prospector seems to be swallowing multiple instances of errors on the same line.

While using South and it's generated migrations, you'll get lot of automatically generated code with poor PEP-8 compliance, so it would be great to ignore them.

It could probably be done automatically once using Django (ignore */migrations/*.py) or providing some generic way to define files to be skipped from checking.

As a supported framework. :)

If I run prospector on this codebase, it hangs:
https://pypi.python.org/packages/source/S/SeriesMarker/SeriesMarker-0.0.3.tar.gz#md5=fc546636a48bda938988fea11bcd266b

Any ideas whats wrong?

please have a look at https://landscape.io/github/fsr-itse/EvaP/88/modules/evap/evaluation/models.py#L231. i'm pretty sure our code is not as broken as prospector suggests :)

crate.io seems to have disappeared (see crateio/crate.io#18) so the badges should be removed from the README

Can prospector be given an option to check single files?

Having a single file mode would make building a prospector powered atom.io linter possible.

installed via pip,
executed via py26 in venv

$ python

Python 2.6.8 (unknown, Mar  9 2014, 22:16:00) 
[GCC 4.2.1 Compatible Apple LLVM 5.0 (clang-500.0.68)] on darwin
Type "help", "copyright", "credits" or "license" for more information.


$ prospector --version                                                                                                                                                                                                         
prospector 0.7

have an error

$ prospector

Traceback (most recent call last):                                                                                                                                                  
  File "/Users/iremizov/Py26/bin/prospector", line 9, in <module>
    load_entry_point('prospector==0.7', 'console_scripts', 'prospector')()
  File "/Users/iremizov/Py26/lib/python2.6/site-packages/prospector/run.py", line 312, in main
    prospector.execute()
  File "/Users/iremizov/Py26/lib/python2.6/site-packages/prospector/run.py", line 253, in execute
    summary['time_taken'] = '%0.2f' % delta.total_seconds()
AttributeError: 'datetime.timedelta' object has no attribute 'total_seconds'

Python supports multiple imports on the same line separated by commas: import os, sys

If someone puts a virtualenvironment in the root of their project, prospector will attempt to check every package. It should detect virtualenvironment directories and skip them

The current README is becoming insufficient to explain all of the features of prospector, and is missing things like the motivation for the project and clarification of some behaviour. A nice big "Read the docs" page would be useful.

The README says to see the supported tools section below, but this is actually another page in the documentation and there should probably be a link to that instead.

The summary is probably a lot more useful if it appears at the end of the output instead of the start.

Prospector and Landscape.io need some way to say "I want 'veryhigh' strictness in general, but use 'medium' for PyLint and use this raw config file with pep8.py instead."

prospector -s medium doesn't enforce PEP8 strictly enough but prospector -s high makes PyLint too pedantic and makes PEP8 complain about "E401: multiple imports on one line", which I've ignored in my flake8 configuration.

As-is, I have to continue to enforce my PEP8 policy by leaving Travis-CI configured to run flake8 and fail the entire test run if the code fails the style check.

For pep8.py, might I suggest just restoring its ability to let a [pep8] section in setup.cfg or tox.ini override the default values? (In other words, supply the prospector configuration via whatever mechanism pep8 --config uses and make sure Prospector isn't confusing its idea of where the project's root directory is.)

https://pypi.python.org/pypi/vulture

for ci-tools designed to parse pylint output

To ignore a specific error/warning, as Prospector is a collection of tools, it could either:

• Show in README.md how to exclude warnings for different tools
• Implement a general prospector warning exclusion (better)

Known exclusion in comment:

• PyLint, usually before a line block: # pylint: disable=MESSAGE_ID_OR_KEY
• PEP-8 add at end of line: # noqa
• dodgy: ?
  -...

Currently, in code written to work in both Python 2.x and Python 3.x, the default strictness_medium profile for prospector causes PyLint to return this error when run under Python 2.x:

run_tests (run_tests.py):
    L41:0 : pylint - C0325
    Unnecessary parens after 'print' keyword

It's hard to make it part of CI because:

• There is no Jenkins plugin that can read the outputs (if there is, please add it to the Wiki).
• The tool doesn't exit code 1 or more when there are lint warnings/errors

I'd suggest first to support exit code.

There seems to be an inconsistency when reporting column numbers: the pylint tool returns 0-based columns, while all other tools report 1-based columns (provided the column number is available, that is):

$ echo import sys >test.py
$ prospector -B -M test.py 
Messages
========

test.py:
    L1:1 None: frosted - E101
     sys imported but unused

test.py:
    L1:1 None: pyflakes - FL0001
    'sys' imported but unused

test (test.py):
    L1:0 None: pylint - W0611
    Unused import sys

Prospector is throwing exception for python package "django-http-proxy". Version of package is 0.3.2.

Below is the stack trace:

Traceback (most recent call last):
  File "/usr/local/bin/prospector", line 9, in <module>
    load_entry_point('prospector==0.6.4', 'console_scripts', 'prospector')()
  File "/usr/local/lib/python3.4/dist-packages/prospector-0.6.4-py3.4.egg/prospector/run.py", line 190, in main
    prospector = Prospector(config, paths[0])
  File "/usr/local/lib/python3.4/dist-packages/prospector-0.6.4-py3.4.egg/prospector/run.py", line 34, in __init__
    self._determine_adapters()
  File "/usr/local/lib/python3.4/dist-packages/prospector-0.6.4-py3.4.egg/prospector/run.py", line 46, in _determine_adapters
    for name, adaptor in autodetect_libraries(self.path):
  File "/usr/local/lib/python3.4/dist-packages/prospector-0.6.4-py3.4.egg/prospector/autodetect.py", line 69, in autodetect_libraries
    adaptor_names = find_from_path(path)
  File "/usr/local/lib/python3.4/dist-packages/prospector-0.6.4-py3.4.egg/prospector/autodetect.py", line 35, in find_from_path
    names |= find_from_path(item_path)
  File "/usr/local/lib/python3.4/dist-packages/prospector-0.6.4-py3.4.egg/prospector/autodetect.py", line 35, in find_from_path
    names |= find_from_path(item_path)
  File "/usr/local/lib/python3.4/dist-packages/prospector-0.6.4-py3.4.egg/prospector/autodetect.py", line 38, in find_from_path
    names |= find_from_imports(fip.read())
  File "/usr/lib/python3.4/codecs.py", line 313, in decode
    (result, consumed) = self._buffer_decode(data, self.errors, final)
UnicodeDecodeError: 'utf-8' codec can't decode byte 0x89 in position 37: invalid start byte

Please let me know if you need any further details to look into this issue..

Could you please look into this ?

Update the documentation to include instructions on "how to get the best out of prospector" namely, install in a virtualenv alongside the project.

The change to use jayclassless/setopfconf removed the ability to specify the path from a command line option. This may be used by people already, so should be reinstated for backwards compatability (and possibly deprecated).

See #14

If you put the line number after the file, then will be much more useful to emacs users that use the compile command. This:

foo/__init__.py :
    L5:1 None: pyflakes - FL0001
    'ApiException' imported but unused

should be:

foo/__init__.py:5 :
    L5:1 None: pyflakes - FL0001
    'ApiException' imported but unused

in https://landscape.io/github/fsr-itse/EvaP/95/modules/evap/evaluation/models.py#L330 i got five warnings for two disallowed spaces. one would have been enough :) maybe two would be acceptable (because there were two spaces disallowed), but not more please.

In order to add Prospector to my Ohloh "tools I actively use for development" list, I took advantage of their "you don't have to be a project manager to add an entry" feature. (It's also one more way people can discover Prospector and Landscape.io)

You'll probably want to declare yourself a project manager on this entry. (Among other things, it gives you the option of disallowing edits by random Joe Nobodies like me.)

https://www.ohloh.net/p/landscapeio-prospector

I've filled out everything I could but I highly recommend checking these pages to see if there's anything I missed:

• News Feeds (Ohloh lets you incorporate announcements into their system via RSS)
• Links (In case there are any categories you can add which I couldn't)
• Logo (In some places, "stuff I use" lists are just icon grids and it's currently using a placeholder)
• Aliases (In case you need to claim commits under multiple names)

Finally, while I think they're large and gaudy, Ohloh does have their own statistics widgets. (But only for GitHub Pages. They use <script> tags)

Sometimes, as seen here: landscapeio/landscape-issues#36 , the sensible defaults are not helpful. There are several inclusions and exclusion rules that are essentially hardcoded. It'd be useful to expose all of these for edge cases (or rather, cases I forgot to think about!)

For this code:

from twisted.internet.protocol import DatagramProtocol

class Class(DatagramProtocol):
    def __init__(self, string, host, port):
        self.transport.write(string, (host, port))

The inspector gives an error that class Class has no method transport. Whereas transport is defined by AbstractDatagramProtocol which is inherited by DatagramProtocol

The radon project calculates the maintainability index of each python file.
It's a very useful metric to understand which modules are way too complex and need to be broken down into smaller modules.
We can use radon programmatically in order to calculate it.

https://pypi.python.org/pypi/pep257/0.3.2

There is no such flag, right?

This may be a little pathological but it's worth knowing about at least: if prospector is installed on the system, then installed into an activated virtualenv, the system version is used. If the virtualenv is deactivated then reactivated, the virtualenv version is used.

When Prospector is run on some python libraries, it simply exit out with message:

"Segmentation fault (core dumped)"

Sample libraries: pygtkspellcheck, Netzob, GladeBuilder, gtimelog

Please let me know if you need any further information.

There is no option to alter the Python path used to look for modules during the static analysis, thus causing a lot of "pylint - F0401 - Unable to import" error message.

The 1.1 release of pylint removed some checkers which prospector uses.

This is somewhat related to pre-commit/pre-commit#178

I'm working on setting up a Prospector hook for http://pre-commit.com/ (a handy git pre-commit hook manager). The way I'm configuring it, pre-commit runs prospector via xargs. See https://github.com/pre-commit/pre-commit/blob/master/pre_commit/languages/system.py

I'm currently setting the entry point to prospector as -n 1 prospector so that pre-commit ends up running xargs -0 -n 1 prospector, which then calls prospector once per input file. This ends up creating a lot of redundant summaries and message headers in the output, when I would rather just get the messages themselves. I can supply the -M argument to suppress the summaries, but I still get the Messages header once per file, even when there are no messages for that file.

My request is to add some sort of flag for either suppressing the Messages header specifically, or for more generally controlling output verbosity.

If this seems reasonable, I'd be willing to work on it and submit a pull request. Alternatively, if there's a better way to do this or something obvious I'm missing, I would appreciate any help.

(my efforts so far are at https://github.com/guykisel/prospector-mirror [which could probably use a better project name])

It would be helpful to issue new releases more often so we can try them and send feedback sooner.

for CI tools