Hi,

Two things to consider:

in contentmatcher.py: implement _pattern_is_simple as:

def _pattern_is_simple(self, pattern):
""" A "simple" pattern that can be matched with str.find and doesn't
require a regex engine.
"""
return re.escape(pattern) == pattern

I noticed, on my machine and TravisCI, that the test test_whole_words in the suite test_contentmatcher.py does not succeed. From the error message it appears that some changes have been made in Python36's regular expression library that cause these issues.

Because the input files are opened in binary mode and stdout is opened as text, on Windows when redirecting the output to a file you end up with CR CR LF line breaks, which is incorrect, and most editors show as two line breaks (extra blank line in between every line).

The "proper" fix would be to open the file in text mode, but as you're doing binary stuff on it too, it's easiest just to replace CR LF with LF before outputting (which the text-opened stdout will convert back to CR LF).

Fixed in this commit

On https://pypi.python.org/pypi/pss, there are still two links to the old Bitbucket page. These should be updated to the relevant GitHub wiki pages.

I mostly use pss to search CPython clones, and it frequently gets hits on the searchindex.js file generated by Sphinx.

Would it be worth adding "searchindex.js" to the default exclusion regexes?

Excluding it with "--nojs" is easy enough for my particular use case, but it wouldn't work so well for a web programming project that actually had also had real Javascript files to search.

pss currently ignores python source code in .pyw files. Adding '.pyw' to TYPE_MAP in driver.py fixes the issue:

'py':
    TypeSpec(['.py', '.pyw'], []),
'python':
    TypeSpec(['.py', '.pyw'], []),

The .pyw extension is used in the windows version of python to suppress the console for GUI applications.
http://docs.python.org/3/tutorial/interpreter.html?#executable-python-scripts

I'm running the current eliben/pss master on Windows 7 64-bit, and running test/all_tests.py gives this:

..F.........................
======================================================================
FAIL: test_ignore_dirs (test_filefinder.TestFileFinder)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "C:\Users\Hoyt\Downloads\pss\test\test_filefinder.py", line 120, in test_ignore_dirs
    'simple_filefinder/partialignored/found.c'])
  File "C:\Users\Hoyt\Downloads\pss\test\test_filefinder.py", line 49, in assertPathsEqual
    list(sorted(map(os.path.normpath, second))))
AssertionError: Lists differ: ['simple_filefinder\\a.c', 'si... != ['simple_filefinder\\a.c', 'si...

First list contains 1 additional elements.
First extra element 3:
simple_filefinder\partialignored\thisoneisignored\notfound.c

  ['simple_filefinder\\a.c',
   'simple_filefinder\\c.c',
-  'simple_filefinder\\partialignored\\found.c',
?                                              ^

+  'simple_filefinder\\partialignored\\found.c']
?                                              ^

-  'simple_filefinder\\partialignored\\thisoneisignored\\notfound.c']

----------------------------------------------------------------------
Ran 28 tests in 0.166s

FAILED (failures=1)

What are the differences between pss and silver-searcher/ag and ripgrep?

for example, for a file with content:
a
b
match using pattern a\nb

When running:

pss PATTERN

I expect to search in all files recognized by pss, but it seems that only those with TypeValue.EXTENSION are searched. TypeValue.PATTERN files should also appear.

Currently the pyproject.toml is the new way of configuring Python projects. Also tools like black, ruff and many other tools offer possibility to configure tools behavior by using pyproject.toml. Would it possible to offer similar support for pss? Example -A, -B, -C, --color, --ignore-dir and many others would be nice to have defaults from the toml file.

$ pss 'yapf_cmd = '
./gdc-pyFormat.py
360:yapf_cmd = Ppf.yapf_api.FormatFile

$ pss --smart-case 'yapf_cmd = '
<<unexpected error: '>=' not supported between instances of 'int' and 'str'>>

Same effect whereever I use option '--smart-case', also combined with other otions.

NB: This is my first github-issue, so excuse if not appropriate
Best regards

Migrated from issue #42 on bitbucket

I love pss as a drop-in replacement - who wants to install Perl? But it odesn't support config files yet - it'd be great it if supported the "standard" ackrc.

Thanks for a great clone!

Not sure if this is a deliberate design decision. It's different compared to ack, ripgrep, silver-searcher.

diff --git a/psslib/defaultpssoutputformatter.py b/psslib/defaultpssoutputformatter.py
index 904a6b8..c1ef337 100644
--- a/psslib/defaultpssoutputformatter.py
+++ b/psslib/defaultpssoutputformatter.py
@@ -46,7 +46,8 @@ class DefaultPssOutputFormatter(OutputFormatter):
         self.style_filename = (decode_colorama_color(filename_color_str) or
                                colorama.Fore.MAGENTA + colorama.Style.BRIGHT)
         self.style_lineno = (decode_colorama_color(lineno_color_str) or
-                             colorama.Fore.WHITE)
+                            colorama.Fore.GREEN + colorama.Back.BLACK)
+        #                     colorama.Fore.WHITE)

         # colorama.init is doing something strange on Linux, by emitting a color
         # reset sequence when it shouldn't. This causes strange errors when

is ripgrep-like.

Command-line tools should generally be safe to run with no arguments. Spewing help all over your terminal (and thus hiding your history/interrupting your flow) is quite unsafe. Please change it to just tell the user to use -h for usage in that case.

Hi Eli. Does pss tool support syntax-aware search that can look inside Python strings only? I need to find a specific string value, but unfortunately it is also a rather common word for my codebase. I could parse AST myself, but it is not practical for daily usage.

I'd like to copy-paste pss output elsewhere but line numbers in the beginning of each line are on a way. Easy to remove them afterwards, but using something like --noline would obviously be even easier.

--noline would be consistent with the current --column. For consistency/symmetry reasons it might be good idea to support also --line and --nocolum.

I use pss extensively because I feel comfortable with Python's regex format, which I prefer over the one sed uses.

However, sed is extremely useful for recursive in-place regex replace.

Do you think it would be a good idea to add search-and-replace features to pss, using re.sub?

I think it deserves to be packed as a snap - https://docs.snapcraft.io/python-apps/6741 - to make it installable with snap install pss.

I'm working with custom file extension .robot (test data files for http://robotframework.org) and by default pss doesn't, as designed, look into them. I can simply use -a to look everywhere, but if I want to look only to those files I need to use -aG '.*\.robot'. Not so nice if you are using pss with such files a lot. There are obviously many other custom file types too, for example, .tmpl is mentioned issue #1 comments.

It would be convenient if there was an option to easily limit searching to a certain file extension. Basically I would like to use something like -E robot instead of -aG '.*\.robot'. Notice that I would like both to avoid writing a regexp when I don't really need it (robot vs. '.*\.robot') and would like -a to be implied. (I was rather surprised I needed to use it with -G, might be a good idea to mention that in its docs.)

I'm willing to implement this functionality myself. Mainly wanted to submit an issue first to hear what you think about it before starting to create a pull request. Should also decide what is a good short option for this usage and is long option needed too.

I'm also looking for a way to add these custom files into a list of files that pss looks by default. That's definitely a topic for a separate issue, most likely my needs are already covered by #1.

This is a common problem I wish pss solved.

• Imagine I have a python class instance method in git REPO_01. This repo has several git branches.

• Imagine my python in git REPO_02 has many git branches, only some of them import the class from REPO_01

Assume I git clone both repos in a directory on my laptop; assume the directory is called ~/project_foo.

How can I iterate over and search all branches and all files of REPO_01 and REPO_02 (which have multiple levels of subdirectories) for a method string, "do_useful_thing_01"?

If I am right, pss can't do this today without installing xargs.

I find ack's output with colored line numbers slightly easier to read, so I added support for that via --color-lineno in this commit -- feel free to take it if you like.

The default is unchanged (gray just like the standard pss output). Ideally it'd be bright yellow by default, but I'll leave that up to you.

FWIW, ack's bright green for filenames is much easier to read (on Windows? my machine?) than pss's magenta. But that's already overridable via --color-filename.

It's really time...

While at it, drop support for all older versions of 3.x which aren't supported any more at https://www.python.org/downloads/

I've been using pss for years mostly when developing Robot Framework. It would be convenient if our custom extensions .robot and .resource were recognized by pss automatically. If you consider this ok, I'm happy to provide a PR.

Minified javascript is easy to detect, when search result is on the line whose length is longer than e.g 1000 characters (majority of style guides recommend less than 80) that result is probably from minified file.

Or allow some kind of .pssrc configuration for this.

For example, coloring works as expected with no command-line arguments:

pss foo

But with certain command line arguments pss doesn't do any coloring:

pss --context=1 foo

However, some command line arguments do still color correctly, like pss -a foo.

I'm using pss on Python 2.7.5, Windows 7 64-bit.

Some projects I work on use a bunch of not-very-standard file extensions, such as .tmpl for mixed HTML templates. With ack I could use --type-add html=.tmpl to include this.

It would be great to have this feature in pss. It doesn't need to be permanent, because I always run pss via a customized script with my own options anyway (essentially an ackrc replacement).

One can kind of achieve this behaviour with a combination of -a and -G, but then it only searches .tmpl files, whereas I want to include them in all my searches by default. Basically --type-add would add to TYPE_MAP.

Let me know if you'd like a patch for this. I think it'd be pretty straight-forward to add these options and have them adjust TYPE_MAP.

$ pss "->"
Usage: pss [options] [files]

pss: error: no such option: ->

I'm using pss in a shell script and want to do different actions depending are there matches or not. I can do that based on does pss write anything to stdout:

if [[ -n "$(pss pattern location)" ]]; then 
     echo match
fi

It would be a bit more convenient to do this based on the return value:

if pss patten location > /dev/null; then
    echo match
fi

Because pss returns 0 regardless does it match anything or not, the above always echos "match". I don't know about ack, but at least grep returns 0 if there is a match, 1 if there is no match, and 2 if there is an error. I would be convenient if pss followed these rules.

At some point pss should use the argparse library instead.

Usually I have a file containing all source files of the project (to be used by cscope).
It could be useful if pss can be instructed to use this file as the search space:
For example:
pss --cc --file-list 'file-list.txt' 'malloc'

I have:

$ pss --py -f   # produces many files
$ pss --scons -f # also many files
$ pss --scons --py -f # NO FILES ?!

This could be related to Issue #2

If --include-pattern or --exclude-pattern are given more than once only the last occurrence is taking effect, all others are effectively ignored.

This is sort of counter intuitive. I expected the following invocation:

pss --include-pattern=A --include-pattern=B

to be equivalent to

pss --include-pattern='(A|B)'

I know that since pss supports Python's regular expression syntax this seems like a mere question of preference, but there exist use cases where the combination behavior is desired.

Example
I use vim (neovim) as my editor of choice. Vim supports the grepprg option, that allows to specify an external program to be used for vim's grep command. I prefer to use pss for this, because I am more familiar with PCRE. In addition it is possible to provide the command with command line options that will always be used. Since I develop Qt applications I am dealing with a lot of auto-generated files of the form *moc_.cpp. These files are only cluttering the search results so I want them excluded. The problem is when I provide this exclude pattern as part of the default command invocation it is not possible to use --exclude-pattern again.
So I am left with one of two options:

1. Write --exclude-pattern='moc_.*cpp' everytime I search something
2. Accept that my search results are cluttered when I use another exclude pattern in a search