Add algorithm for testing Project Euler solutions

[dhruvmanila]

As described in #2463 (comment)
Currently in beta

Adding solution_test.py file for testing all the submitted solutions for Project Euler. The file will not be picked up by Travis CI and we will have to manually call it, maybe we can call it in before_script section for Project Euler build.

The use case for imported modules:

• pathlib:
  • Path manipulation
  • Iterating over the solution files path
  • Excluding all non-solution files
• importlib.util
  • For importing the solution files directly using it's path
• logging
  • For logging errors like:
    • File doesn't contain solution() function
    • solution() function requires a positional argument
    • Incorrect answer for the problem

Currently all the logs goes into the filename described in LOG_FILENAME constant, we can change it to wherever we want it to go. The log looks like this (we can change it however we want):

ERROR: solution() missing 1 required positional argument: 'n' 
Source: Problem 1: sol7.py
...

ERROR: module 'sol3.py' has no attribute 'solution' 
Source: Problem 10: sol3.py
...

Describe your change:

• Add an algorithm?

Checklist:

• I have read CONTRIBUTING.md.
• This pull request is all my own work -- I have not plagiarized.
• I know that pull requests will not be merged if they fail the automated tests.
• This PR only changes one algorithm file. To ease review, please open separate PRs for separate algorithms.
• All new Python files are placed inside an existing directory.
• All filenames are in all lowercase characters with no spaces or dashes.
• All functions and variable names follow Python naming conventions.
• All function parameters and return values are annotated with Python type hints.
• All functions have doctests that pass the automated testing.
• All new algorithms have a URL in its comments that points to Wikipedia or other similar explanation.
• If this pull request resolves one or more open issues then the commit message contains Fixes: #{$ISSUE_NO}.

[cclauss]

I put the PR in DRAFT mode. (Kind like beta) ;-)

[dhruvmanila]

Extra ideas:

Should we keep the answers in a separate file in the JSON format? Or maybe extract the answers from https://github.com/nayuki/Project-Euler-solutions/blob/master/Answers.txt every time we call the script using either urllib.request or requests (better option) library?

Do we need additional information in our logs like lineno, asctime ?
All the attributes are present on https://docs.python.org/3/library/logging.html#logrecord-attributes
Should we log with colors?

[cclauss]

This is looking awesome. The logging is just for us to develop the tests. Don't make the logs fancy -- we will get rid of them when we are done. Once we land this PR, files will either pass the test or we will flag them as inadequate.

[dhruvmanila]

Alright. What are you thinking to do next?
How should we implement it into Travis CI? As I mentioned above, we can add python project_euler/solution_test.py to before_script part in the config file or maybe add it in scripts directory as that seems like a place where this script should be.

Once we land this PR, files will either pass the test or we will flag them as inadequate.

So, the tests will fail if the requirements aren't satisfied, printing the appropriate message for the person to debug? I'm a bit confused about what do you mean by "flag them as inadequate"

[cclauss]

Let's run and fail fast just like scripts/validate_filenames.py does...
https://travis-ci.com/github/TheAlgorithms/Python/builds/185558423#L256

[dhruvmanila]

Oh lol, I forgot to change the solution_test.py file. Also, should I change it as to how validate_filename.py is where it groups the files according to the error and print the filenames and exit with appropriate error code?

[cclauss]

So... pytest "test discovery" finds all files called test_xxx.py or xxx_test.py and runs them as tests. So pytest is automatically running the file which means that you do not need to explicitly run the file in .travis.yml.

[cclauss]

It takes 40 seconds to run.
https://travis-ci.com/github/TheAlgorithms/Python/jobs/390526545#L473

[TravisBuddy]

Hey @dhruvmanila,
Something went wrong with the build.

TravisCI finished with status errored, which means the build failed because of something unrelated to the tests, such as a problem with a dependency or the build process itself.

View build log

TravisBuddy Request Identifier: c5c89b70-feac-11ea-b444-3dcc0bd866ab

[dhruvmanila]

This commit is going to fail as for what I wanted to show when you mentioned:

Let's run and fail fast just like scripts/validate_filenames.py does...

https://travis-ci.com/github/TheAlgorithms/Python/jobs/390693925#L249

[TravisBuddy]

Hey @dhruvmanila,
Something went wrong with the build.

TravisCI finished with status errored, which means the build failed because of something unrelated to the tests, such as a problem with a dependency or the build process itself.

View build log

TravisBuddy Request Identifier: 8df1c1d0-fead-11ea-b444-3dcc0bd866ab

[cclauss]

commit is going to fail

That is why we are in draft mode. ;-)

OK... So not how do we read the function signature so that we know how many input parameters a function needs? For each directory, we will need some sample data to pass into those functions that take parameters.

[dhruvmanila]

The input parameters come from the questions. We can have something like if the function requires arguments, which would be in the case of doctest, the default argument(s) should be the question's input parameters.

In the other case, we need to have those input parameters and if required, which we will know if we get a TypeError, we will pass those as the arguments. That will require on our end to have those input parameters.

[cclauss]

from collections import namedtuple

q_and_a = namedtuple("q_and_a", "question answer")

ANSWERS = {
1: q_and_a(None, "233168"),
2: q_and_a(1, "4613732"),
3: q_and_a((7, 3), "6857"),
4: q_and_a("Mary", "906609"),
...
}

Just do a few to see if we can make it work.

[dhruvmanila]

Not possible. Just take this problem as an example: https://projecteuler.net/problem=4

Here they are asking for the largest palindrome made from the product of two 3-digit numbers. Now, how do we define an input parameter to this question? The solutions submitted have n as a number up to which it's going to look to find the largest palindrome product, but the question is asking for the largest which, without knowing the answer, we have no idea what the limit will be. n could have been n-digit number up to which to find the largest palindrome number and then for this question n would be 3.

We do know the answer but that's not my point. We cannot define an input parameter to such questions. The solution() should define the default input parameters according to the question and the solution they came up with such that when we call the function with no arguments it defaults to the actual answer.

Just take a look at the size of this input parameter lol: https://projecteuler.net/problem=13

[cclauss]

If solution() needs no parameter then question=None like I did with number 1
If solution() needs one parameter then question = one parameter like I did with number 2
if solution() needs multiple parameters then question = a tuple of parameters like number 3

[dhruvmanila]

So you're saying we provide the parameters? What do you suggest the parameter for https://projecteuler.net/problem=4 be? The solutions submitted does ask for a parameter.

[dhruvmanila]

Hi @cclauss, Can you take a look at this? Thanks! 😁

Sorry, it was pretty late so I went to sleep.
Here you can see that some of the problem input get really large and a lot of are None. This is pretty inefficient. I would highly suggest, for the long run, to enforce either no parameters for solution() or to have the default parameters as the question's input which will be according to their solution and the given problem.

https://travis-ci.com/github/TheAlgorithms/Python/jobs/390761149#L243
As you can see it's taking double time now, also some of the functions are returning list or dict object.

Take a look at: https://github.com/nayuki/Project-Euler-solutions/blob/master/python/eulertest.py#L21
It's one function which directly returns the answer, making life so much easier for testing ;-)

[TravisBuddy]

Hey @dhruvmanila,
Something went wrong with the build.

TravisCI finished with status errored, which means the build failed because of something unrelated to the tests, such as a problem with a dependency or the build process itself.

View build log

TravisBuddy Request Identifier: 83621750-ff03-11ea-b444-3dcc0bd866ab

[cclauss]

OK... Let's force default args so that solution() can always be called with no args.

On eulertest.py#L21, my modification would be:

• actualans = module.compute() --> actual_answer = str(module.compute())
  This allows modules to return whatever datatype they want but actual_answer is always str.

If it would help, send me an email so we can do a videoconf.

[dhruvmanila]

Alright, I did it and also added all the answers according to https://github.com/luckytoilet/projecteuler-solutions

actualans = module.compute() --> actual_answer = str(module.compute())
This allows modules to return whatever datatype they want but actual_answer is always str.

I already did this right from the start 😁

I'll send you the email but is there anything else to discuss?

[TravisBuddy]

Hey @dhruvmanila,
Something went wrong with the build.

TravisCI finished with status errored, which means the build failed because of something unrelated to the tests, such as a problem with a dependency or the build process itself.

View build log

TravisBuddy Request Identifier: 01f25b50-ffe7-11ea-9882-0125ab0f75bf

[dhruvmanila]

Alright, that's not so good in terms of information plus it went on executing the other script, I will put it in before_script section. This is reducing the overall time by a large margin, from 60-80s to 30s.

Any idea about how to capture just the message part of the exception in the pytest and exiting with failure?
We could use except block and print the message like before but if we handle the exception pytest will think the function passed the test.

[TravisBuddy]

Hey @dhruvmanila,
Something went wrong with the build.

TravisCI finished with status errored, which means the build failed because of something unrelated to the tests, such as a problem with a dependency or the build process itself.

View build log

TravisBuddy Request Identifier: e2767590-00ef-11eb-a257-6fde496095bb

[dhruvmanila]

I fixed long error messages into one line like before and now Travis CI will exit if this script fails.
@cclauss Can you take a look at this? And what is hacktoberfest?

[dhruvmanila]

I removed the patch for printing the error messages and using fixture to do it.
Source: https://stackoverflow.com/a/52873379

That's weird. Why did Travis start two builds?

[cclauss]

Awesome work @dhruvmanila Congratulations! 💯

@poyea @AnupKumarPanwar @dynamitechetan This is an awesome submission made by @dhruvmanila. Well written, modern Python which delivers the superpower of validating the answer to every Euler solution in our repository. It scans every subdirectory and finds and runs the proposed Euler solution and then lists out any algorithm that does not return the correct answer.

I would like to propose that we add @dhruvmanila as a maintainer of TheAlgorithms.

[dhruvmanila]

@cclauss Thank you for giving me the opportunity to work on it. 😇