Use absolute file paths in ITPParser by aditya-kamath · Pull Request #3108 · MDAnalysis/mdanalysis

aditya-kamath · 2021-01-24T17:53:30Z

Fixes #3037

Changes made in this Pull Request

ITPParser now accepts relative paths

PR Checklist

[] Tests?
[] Docs?
CHANGELOG updated?
Issue raised/referenced?

richardjgowers

Needs a test

IAlibay

Couple of extra things on top of @richardjgowers' comment

package/CHANGELOG

package/MDAnalysis/topology/ITPParser.py

IAlibay · 2021-01-24T18:12:45Z

This might need some rethinking, it's breaking CI: https://github.com/MDAnalysis/mdanalysis/pull/3108/checks?check_run_id=1757898726#step:10:999

aditya-kamath · 2021-01-24T18:18:28Z

Hmm, I guess I should understand how these path strings are manipulated in general. This did seem to work when I tried to load a .ITP file from another folder.

tylerjereddy · 2021-01-25T00:25:23Z

package/MDAnalysis/topology/ITPParser.py

@@ -268,7 +268,7 @@ def find_path(self, path):
            path = path.name


A slightly different change seems to allow the testsuite to pass while still allowing relative file path handling in my hands:

--- a/package/MDAnalysis/topology/ITPParser.py +++ b/package/MDAnalysis/topology/ITPParser.py @@ -265,7 +265,7 @@ class GmxTopIterator: current_file = self.current_file try: - path = path.name + path = os.path.abspath(path.name) except AttributeError: pass

I think it addresses Irfan's comment as well

codecov · 2021-01-25T16:20:25Z

Codecov Report

Merging #3108 (84d0fd4) into develop (8a5ef48) will not change coverage.
The diff coverage is 100.00%.

@@           Coverage Diff            @@
##           develop    #3108   +/-   ##
========================================
  Coverage    92.81%   92.81%           
========================================
  Files          170      170           
  Lines        22801    22801           
  Branches      3239     3239           
========================================
  Hits         21162    21162           
  Misses        1591     1591           
  Partials        48       48

Impacted Files	Coverage Δ
package/MDAnalysis/topology/ITPParser.py	`96.84% <100.00%> (ø)`

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8a5ef48...84d0fd4. Read the comment docs.

aditya-kamath · 2021-01-28T00:10:25Z

@richardjgowers, To add a test, would it be okay to make another folder and show that it can access an ITP file? I'm not able to figure out a better test..

IAlibay · 2021-03-10T17:04:38Z

@MDAnalysis/gsoc-mentors can we get a follow-up review here?

@aditya-kamath can you fix the merge conflict and also maybe provide some example code for what you mean in #3108 (comment)? (that way it's easier for us to comment on how best to change it).

aditya-kamath · 2021-03-10T22:27:00Z

I guess I'm just asking what kind of test should I provide in this case? I want to show that this parser can now access paths outside root directory. I tried this manually and it works but I'm thinking in a pytest context. :)

hmacdope · 2021-03-10T22:33:52Z

In a pytest context you could try just providing any relative path to the ITP you are meant to be testing. This includes the ITPs in the /MDAnalysisTests/data/ folder. That's probably a good place to start.

pep8speaks · 2021-03-11T20:19:30Z

Hello @aditya-kamath! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2021-04-15 21:32:36 UTC

…lysis into ITPParserfix

aditya-kamath · 2021-03-11T22:44:21Z

I've added a test, and it seems to work fine when I do pytest test_itp.py. But, this seems to fail? I've also checked that it does indeed work with relative paths.

IAlibay

I've added a test, and it seems to work fine when I do pytest test_itp.py. But, this seems to fail? I've also checked that it does indeed work with relative paths.

See here: https://github.com/MDAnalysis/mdanalysis/pull/3108/checks?check_run_id=2089681447#step:10:366

It doesn't seem like you're creating the temporary ITP file?

IAlibay · 2021-03-11T22:48:18Z

testsuite/MDAnalysisTests/topology/test_itp.py

+class Testrelativepath:
+
+    def test_relpath(self):
+        u = mda.Universe('../data/gromacs_ala10.itp', format='ITP')


Where is this itp file being generated?

Okay, I get it. I guess I need to load this file first..!

@aditya-kamath Python treats files with relative paths, as relative to whichever directory you've called the test from. Users can call pytest from anywhere in their system, which is why we make all our paths absolute in the datafiles.py file and import them when we want to use our test files. It might be easier for you to write a temporary ITP file using pytest's tools that only exists for the scope of this function or module, and try to read that.

@lilyminium , This really helps. Now, I understand how to write this test! :)

I've now created a temporary folder where an itp file is written and have accessed it with the parser. Although, I am still not completely sure if this asserts that itp parser can access relative locations.?

lilyminium

Thank you for fixing this, @aditya-kamath. I just have a few comments :)

lilyminium · 2021-03-27T00:58:55Z

testsuite/MDAnalysisTests/topology/test_itp.py

            with self.parser(ITP_no_endif) as p:
                top = p.parse(include_dir=GMX_DIR)
+
+class Testrelativepath:


Please follow Python convention and use camel casing here if you create a class. However, since you only have one tests, I would suggest that you don't need a class for this test. Instead, you could structure test_relpath as a standalone function.

lilyminium · 2021-03-27T01:00:12Z

testsuite/MDAnalysisTests/topology/test_itp.py

+class Testrelativepath:
+
+    def test_relpath(self, tmp_path):
+        content="[ atoms ]\n" +\


It would be neater here to use """ multiline strings. I don't think the indentation will matter to the ITPParser.

lilyminium · 2021-03-27T01:03:29Z

testsuite/MDAnalysisTests/topology/test_itp.py

+        d.mkdir()
+        p = d / "test.itp"
+        p.write_text(content)
+        u = mda.Universe(tmp_path / "sub/test.itp", format='ITP')


I think this would still be an absolute path. Instead, I would suggest using the tmpdir fixture, creating a new directory inside that temporary one, and making sure that the relative path you test looks something like "../test.itp".

Hi @lilyminium, I've been looking at pytest tools and os.paths. I see that there's a command called relative_to in the pathlib module which computes a relative path to a specified directory. My approach would have been to make a tmpdir and set that as 'observer' directory and use that command (relative_to) to find the path that it sees.

As it turns out, the relative_to command only works when computing paths from parent directory reference, not sibling directory. So, I can't compute something like '../dir'. The second thing is even if I manually create a relative string which tmpdir would use to read my temporary itp file, I am not sure how to ask pytest to run from my tmpdir while performing this test.

I might have missed something here, especially because I am new to pytest, but I am happy to hear any advice you have here!

Hi @aditya-kamath,

If you go into your temporary directory, write a file called "test.itp", make a new subdirectory, and change into that directory, you should be able to test that file with the path "../test.itp". You might like the tmpdir.as_cwd context manager for that.

Edit: see this StackOverflow question for examples and notes https://stackoverflow.com/questions/55658866/how-to-get-the-temporary-path-using-pytest-tmpdir-as-cwd

Great advice! I figured the tests out. Just a note that I've done the reverse of what you might expect. Instead of testing a relative path and then converting it to a relative string. I've first tested a relative string and then followed it up by converting it to a path. Hope that's alright!

lilyminium · 2021-03-27T01:07:59Z

package/MDAnalysis/topology/ITPParser.py


        try:
-            path = path.name
+            path = os.path.abspath(path.name)


This try/except is to check whether path is a string (which triggers the AttributeError) or a Path object from pathlib (so Path.path is the string version). Does your fix still work if a relative string path is passed in instead, triggering the AttributeError?

By the way, your test should check for both, just in case :)

Hi @lilyminium, that's a good observation! However, if the user isn't providing a string and triggering this AttributeError, it's probably doing the right thing by just passing and eventually raising an IO error. I can write a test to check if an IOerror is raised when a relative string is passed but not as a string.

Hi @aditya-kamath,

if the user isn't providing a string and triggering this AttributeError

If a string is provided and the AttributeError is triggered, does your fix still work? The IOError test is probably not necessary. However, your test should check that your fix applies to both a relative Path (https://docs.python.org/3/library/pathlib.html) and a relative string.

(so Path.path is the string version)

This was a typo, sorry. Path.name returns the base name of the Path, as a string.

Hmm, the answer to if it will work when there's an Attribute error is no! :(
I'm assuming that the situation when this will happen is when os.path.abspath is not able to convert a relative string to an absolute path. I'll write the two situations in my test to see if there's a case when this might happen! :D

lilyminium · 2021-03-27T01:08:15Z

package/CHANGELOG

  * 2.0.0

 Fixes
+  * ITPParser now accepts relative paths (Issue #3037)


Suggested change

* ITPParser now accepts relative paths (Issue #3037)

* ITPParser now accepts relative paths (Issue #3037, PR #3108)

Thanks to Lilyminium!

lilyminium · 2021-03-31T00:57:26Z

Thanks for adding the new tests, @aditya-kamath. It's unexpected that they pass, as I expect not accounting for the AttributeError to break the test. On further investigation, the ITPParser could always deal with relative paths when used to create a Universe. I tested this by reverting your os.path.abspath change; the tests still pass. However, what it cannot deal with is relative paths when they're #included in earlier ITPs. That means the test case should actually be a file like this:

#include "../sub1/atoms.itp"
        
[ atoms ]
 1      H      1    SOL    HW1      1       0.41    1.00800

with a corresponding atoms.itp. This means that instead of writing a temporary file, you could actually just include a new ITP file in the data directory and as a variable in datafiles.py that references an upper-level ITP file. It's up to you which one you choose.

@richardjgowers did you first find the bug when one of your ITP files #included an upper level ITP file, or was the relative path included in some other way?

aditya-kamath · 2021-04-05T17:18:25Z

Okay, sounds good! I'll try to get the parser working for all these cases. The next steps may be to emit warnings for broken files...

aditya-kamath · 2021-04-10T03:10:25Z

Btw, @lilyminium , I still believe this fix is working well as is. I tested it to see if it could parse files which #included other files and while it could not do it before, it seems to be able to parse relative paths which '#include other paths. I've written a test for this which creates three temporary folders and while located in one of the folders, tries to parse an itp file in it's sibling directory which #includes a second itp file. I've run pytest fine with this but it's breaking CI so I guess I missed something..

lilyminium · 2021-04-10T03:57:13Z

@aditya-kamath your tests still pass even without your changes. It's often good to come up with a falsifying example to start off with. Can you get something like this test to pass?

def test_relative_path(self, tmpdir):
        test_itp_content = '#include "../atoms.itp"'
        atoms_itp_content = """
        [ moleculetype ]
        UNK 3

        [ atoms ]
        1      H      1    SOL    HW1      1       0.41    1.00800
        """

        with tmpdir.as_cwd():
            with open("atoms.itp", "w") as f:
                f.write(atoms_itp_content)
            subdir = tmpdir.mkdir("subdir")
            with subdir.as_cwd():
                with open("test.itp", "w") as f:
                    f.write(test_itp_content)
                subsubdir = subdir.mkdir("subsubdir")
                with subsubdir.as_cwd():
                    u = mda.Universe("../test.itp")
                    assert len(u.atoms) == 1

Having a look at how the function gets the output dir_path may help.

aditya-kamath · 2021-04-10T21:29:31Z

Thanks for the test @lilyminium . It seems to pass on my computer, I've added it to this branch so let's see if it passes the checks here.

aditya-kamath · 2021-04-11T15:49:42Z

All checks passed with the new test, there might yet be a case situation we are missing!

lilyminium · 2021-04-15T18:40:58Z

Awesome, thank you @aditya-kamath -- I think this fix covers most cases then! Could you please just address the PEP8 comments by the bot in #3108 (comment) and then I think this is good to go!

Sorry for the delay in replying, got caught up in other things. Please just ping me when you're happy.

aditya-kamath · 2021-04-15T21:28:11Z

Great! :D
I've fixed most of the PEP8 errors. The last couple of them is because I've used a class to put the test functions in. This is not really necessary, mostly for an organized code structure.
@lilyminium

testsuite/MDAnalysisTests/topology/test_itp.py

lilyminium

The bot is just asking you to add an extra line / remove some lines in the class :)

Co-authored-by: Lily Wang <31115101+lilyminium@users.noreply.github.com>

aditya-kamath · 2021-04-15T22:42:04Z

It's looking good @lilyminium!

lilyminium · 2021-04-15T22:48:03Z

@richardjgowers @IAlibay could you guys please have another look at this? If not, I'll dismiss the reviews as stale.

Dismissing old review as stale.

lilyminium · 2021-04-23T11:30:30Z

Thank you @aditya-kamath for your patience on this on this very helpful fix!

Fixes MDAnalysis#3037 Co-authored-by: Lily Wang <31115101+lilyminium@users.noreply.github.com>

* added get_connections * modified tests for atoms.bonds/angles/dihedrals etc * modified parsers and things to use get_connections or bonds * updated CHANGELOG * pep8 * undo half of PR 3160 * add intra stuff * Update package/MDAnalysis/core/groups.py Co-authored-by: Jonathan Barnoud <jonathan@barnoud.net> * tighten up base class checking * update docstring * suppres warnings * Use absolute file paths in ITPParser (#3108) Fixes #3037 Co-authored-by: Lily Wang <31115101+lilyminium@users.noreply.github.com> * Adds aromaticity and Gasteiger charges guessers (#2926) Towards #2468 ## Work done in this PR * Add aromaticity and Gasteiger charges guessers which work via the RDKIT converter. * BLD: handle gcc on MacOS (#3234) Fixes #3109 ## Work done in this PR * gracefully handle the case where `gcc` toolchain in use on MacOS has been built from source using `clang` by `spack` (so it really is `gcc` in use, not `clang`) ## Notes * we could try to add regression testing, but a few problems: - `using_clang()` is inside `setup.py`, which probably can't be safely imported because it has unguarded statements/ code blocks that run right away - testing build issues is typically tricky with mocking, etc. (though in this case, probably just need to move `using_clang()` somewhere else and then test it against a variety of compiler metadata strings * Remove ParmEd Timestep writing "support" (#3240) Fixes #3031 * Adding py3.9 to gh actions CI matrix (#3245) * Fixes #2974 * Python 3.9 officially supported * Add Python 3.9 to testing matrix * Adds macOS CI entry, formalises 3.9 support * fix changelog * special metaclass * move function down * tidy code Co-authored-by: Jonathan Barnoud <jonathan@barnoud.net> Co-authored-by: Aditya Kamath <48089312+aditya-kamath@users.noreply.github.com> Co-authored-by: Cédric Bouysset <bouysset.cedric@gmail.com> Co-authored-by: Tyler Reddy <tyler.je.reddy@gmail.com> Co-authored-by: Irfan Alibay <IAlibay@users.noreply.github.com>

aditya-kamath added 2 commits January 24, 2021 12:39

abspath call to find_path

7f3af66

Update CHANGELOG

eb6b8aa

richardjgowers previously requested changes Jan 24, 2021

View reviewed changes

IAlibay requested changes Jan 24, 2021

View reviewed changes

package/CHANGELOG Outdated Show resolved Hide resolved

package/MDAnalysis/topology/ITPParser.py Outdated Show resolved Hide resolved

Update changelog

45fa573

tylerjereddy reviewed Jan 25, 2021

View reviewed changes

Slight change to ITPParser

26806fc

IAlibay added the GSOC Starter label Mar 10, 2021

Merge branch 'develop' into ITPParserfix

ba1e31d

test for itpparsefix

9c9ed50

aditya-kamath and others added 4 commits March 11, 2021 15:28

Merge branch 'develop' into ITPParserfix

f6228a4

fix PEP8 issues

62307b6

Merge branch 'ITPParserfix' of https://github.com/aditya-kamath/mdana…

0708e65

…lysis into ITPParserfix

PEP8 fix

a69ecff

IAlibay previously requested changes Mar 11, 2021

View reviewed changes

aditya-kamath added 2 commits March 14, 2021 21:00

Create temporary directory when testing

c7b2fb0

Fix PEP8 errors

16cad3d

lilyminium requested changes Mar 27, 2021

View reviewed changes

aditya-kamath and others added 2 commits March 30, 2021 19:39

Fixed tests to demonstrate relative path and relative string

dda1efa

Thanks to Lilyminium!

Merge branch 'develop' into ITPParserfix

3f12b43

IAlibay added this to the 2.0 milestone Apr 6, 2021

aditya-kamath and others added 2 commits April 7, 2021 16:43

Update test_itp.py

4d7fe44

Merge branch 'develop' into ITPParserfix

54f90d1

Adding another test!

abc9bb3

Merge branch 'develop' into ITPParserfix

34222e5

Correct PEP8 errors

3a1b2b5

lilyminium reviewed Apr 15, 2021

View reviewed changes

testsuite/MDAnalysisTests/topology/test_itp.py Outdated Show resolved Hide resolved

lilyminium reviewed Apr 15, 2021

View reviewed changes

testsuite/MDAnalysisTests/topology/test_itp.py Show resolved Hide resolved

lilyminium requested changes Apr 15, 2021

View reviewed changes

aditya-kamath and others added 2 commits April 15, 2021 17:32

minor pep8

b14e1e5

Co-authored-by: Lily Wang <31115101+lilyminium@users.noreply.github.com>

minor pep8 2

84d0fd4

Co-authored-by: Lily Wang <31115101+lilyminium@users.noreply.github.com>

lilyminium approved these changes Apr 15, 2021

View reviewed changes

lilyminium changed the title ~~Itp parserfix~~ Use absolute file paths in ITPParser Apr 23, 2021

lilyminium merged commit b82635a into MDAnalysis:develop Apr 23, 2021

lilyminium added a commit to lilyminium/mdanalysis that referenced this pull request Apr 28, 2021

Use absolute file paths in ITPParser (MDAnalysis#3108)

3a34048

Fixes MDAnalysis#3037 Co-authored-by: Lily Wang <31115101+lilyminium@users.noreply.github.com>

fiona-naughton added defect Component-Topology labels Sep 26, 2023

		@@ -268,7 +268,7 @@ def find_path(self, path):
		path = path.name

	* ITPParser now accepts relative paths (Issue #3037)
	* ITPParser now accepts relative paths (Issue #3037, PR #3108)

Conversation

aditya-kamath commented Jan 24, 2021 • edited by IAlibay Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

PR Checklist

Uh oh!

richardjgowers left a comment

Choose a reason for hiding this comment

Uh oh!

IAlibay left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

IAlibay commented Jan 24, 2021

Uh oh!

aditya-kamath commented Jan 24, 2021

Uh oh!

Choose a reason for hiding this comment

Uh oh!

codecov bot commented Jan 25, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

aditya-kamath commented Jan 28, 2021

Uh oh!

IAlibay commented Mar 10, 2021

Uh oh!

aditya-kamath commented Mar 10, 2021

Uh oh!

hmacdope commented Mar 10, 2021

Uh oh!

pep8speaks commented Mar 11, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Comment last updated at 2021-04-15 21:32:36 UTC

Uh oh!

aditya-kamath commented Mar 11, 2021

Uh oh!

IAlibay left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lilyminium left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

aditya-kamath Mar 29, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lilyminium Mar 30, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

aditya-kamath Mar 27, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

aditya-kamath commented Jan 24, 2021 •

edited by IAlibay

Loading

codecov bot commented Jan 25, 2021 •

edited

Loading

pep8speaks commented Mar 11, 2021 •

edited

Loading

aditya-kamath Mar 29, 2021 •

edited

Loading

lilyminium Mar 30, 2021 •

edited

Loading

aditya-kamath Mar 27, 2021 •

edited

Loading

aditya-kamath commented Apr 15, 2021 •

edited

Loading

lilyminium commented Apr 15, 2021 •

edited

Loading