Some tests of the Phone Numbers exercise ignoring some requirments for phone numbers

[borderite]

According to the problem description of "Phone Numbers", the format of a phone number is usually represented as

NXX NXX-XXXX

where N is any digit from 2 through 9 and X is any digit from 0 through 9 (when no country code is present). The following tests seem to ignore this rule:

• cleans-numbers-with-dots-test
• valid-when-11-digits-and-first-is-1-test
• area-code-test
• pprint-test
• pprint-full-us-phone-number-test

If the area-code and pprint functions are purely for formatting a phone number in particular ways, the last three tests in the list are OK. But what these functions are supposed to do need to be described in some way (e.g., adding docstrings to these functions templates).

Do you think that I should prepare PR for this? If so, please let me know what the area-code and pprint functions are supposed to do.

[BNAndras]

cleans-numbers-with-dots-test and valid-when-11-digits-and-first-is-1-test aren't using the canonical input / output values so those should be fixed. I'd also want to double-check any canonical tests not already marked as not included in metadata.toml are in fact present.

The other three tests are track-specific and not from the problem-specifications repo. area-code appears to be for just retrieving the area code and pprint is for pretty-printing the cleaned phone number. I feel pprint contradicts the instructions since we list several inputs with the same output of "6139950253". However, numbers and pprint would give different outputs so the instructions are misleading. An instructions append could clarify that pprint gives a different formatted output from numbers.

[borderite]

OK, it sounds that the tasks are beyond my ability. I hope that someone will take them up.

[BNAndras]

You don't have to do everything I mentioned. Even a smaller contribution is helpful. How do you feel about updating the expected inputs and outputs for cleans-numbers-with-dots-test and valid-when-11-digits-and-first-is-1-test? Refer to https://github.com/exercism/problem-specifications/blob/main/exercises/phone-number/canonical-data.json for those values. Let me know if you have any questions, and we can iterate together towards a fix.

[borderite]

Because the tests currently being used look pretty strange, I went over the file exercises/practice/phone-number/.meta/tests.toml. If I understand the file correctly, 14 tests are missing in phone-number-test.el (relative to the canonical test data). In addition, two among the existing tests in phone-number-test.el do not match the corresponding ones in the canonical test data. My impression is that the tests are generated semi-automatically. I am quite puzzled by my findings.

[borderite]

As BNAndras suggested, I have revised the tests deviating from the canonical test data. I also replaced impossible phone numbers with valid ones in the three tests specific to emacs lisp. Then I have sent a pull request for this change.

I can easily continue bringing tests in the canonical test data into the tests of "Phone Number". Any opnion? Also, can anyone tell me how to add a track-specific description to the problem description?

[BNAndras]

See https://exercism.org/docs/building/tracks/practice-exercises#h-documentation-files. We’d want an instructions append, which the site adds to the end of the instructions file for that exercise. The Anagram exercise should have a recent append you can look at.

[fapdash]

@borderite Example append: https://github.com/exercism/emacs-lisp/blob/b61efbba40c05903ba79525345a2e4b894b7598e/exercises/practice/anagram/.docs/instructions.append.md?plain=1
Note that the level 1 headline will be ignored, but level 2 and below headlines will show up in the web ui.

14 tests are missing in phone-number-test.el (relative to the canonical test data)

I think there are 10 tests missing.
The canonical-data.json contains 22 test definitions, 4 of the are reimplementing / overwriting old test cases, leaves us with 18.
And we have 8 tests, when we exclude the tests for the functions specific to the Emacs Lisp track. The confusing thing is that the tests are not following the naming pattern of the canonical data and they don't follow the order of the tests in the canonical data, which makes it hard to cross reference.

You can use bin/generate_practice_exercise to generate stubs for the tests. The stubs will include the canonical test definitions as comments for easy reference, but it won't generate the actual test code.

I would suggest the following workflow:

1. rename exercises/phone-numbers to exercises/phone-numbers.bak
2. run bin/generate_practice_exercise phone-numbers
3. copy over the test implementations of the tests that where already implemented from exercises/phone-numbers.bak. now their names match the names in the canonical data (description field in the canonical-data.json)
4. implement the new tests

[borderite]

@fapdash Thanks for the information and suggestions. I will work on the remaining tests soon.

[fapdash]

@borderite Great, thanks a lot! :)

[borderite]

@fapdash I followed the steps you described earlier to generate a template test file. But it failed saying the same problem was already in the config file. After removing the entry of "Phone Number" from the config file, I successfully got the template file and created the test file. Now, I have a question. When I ran bin/generate_practice_exercise phone-numbers, it put a new entry in the config file with a new uuid. Should I keep the modified config file, or should I copy the original one back to it?
Thanks

[borderite]

I have revised the test file using custom errors instead of errors messages.