gergelypolonkai-web-jekyll/_posts/2018-07-26-check-if-last-git-commit-has-test-coverage.md

215 lines
7.0 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

---
layout: post
title: "Check if the last Git commit has test coverage"
date: 2018-07-26 12:49:52
tags: [python,development,testing]
published: true
author:
name: Gergely Polonkai
email: gergely@polonkai.eu
---
I use Python at work and for private projects. I also aim to write tests for my code, especially
recently. And as I usually dont start from 100% code coverage (TDD is not my game), I at least
want to know if the code I just wrote have full coverage.
The trick is to collect all the lines that changed, and all the lines that has no coverage. Then
compare the two, and you have the uncovered lines that changed!
### Getting the list of changed lines
Recently, I bumped into
[this article](https://adam.younglogic.com/2018/07/testing-patch-has-test/). It is a great awk
script that lists the lines that changed in the latest commit. I have really no problem with awk,
but Im pretty sure it can be done in Python, as that is my main language nowadays.
```python
def get_changed_lines():
"""Get the line numbers that changed in the last commit
"""
git_output = subprocess.check_output('git show', shell=True).decode('utf-8')
current_file = None
lines = {}
left = 0
right = 0
for line in git_output.split('\n'):
match = re.match(r'^@@ -([0-9]+),[0-9]+ [+]([0-9]+),[0-9]+ @@', line)
if match:
left = int(match.groups()[0])
right = int(match.groups()[1])
continue
if re.match(r'^\+\+\+', line):
current_file = line[6:]
continue
if re.match(r'^-', line):
left += 1
continue
if re.match(r'^[+]', line):
# Save this line number as changed
lines.setdefault(current_file, [])
lines[current_file].append(right)
right += 1
continue
left += 1
right += 1
return lines
```
OK, not as short as the awk script, but works just fine.
### Getting the uncovered lines
Coverage.py can list the uncovered lines with `coverage report --show-missing`. For Calendar.social, this looks something like this:
```
Name Stmts Miss Cover Missing
----------------------------------------------------------------------
calsocial/__init__.py 173 62 64% 44, 138-148, 200, 239-253, 261-280, 288-295, 308-309, 324-346, 354-363
calsocial/__main__.py 3 3 0% 4-9
calsocial/account.py 108 51 53% 85-97, 105-112, 125, 130-137, 148-160, 169-175, 184-200, 209-212, 221-234
calsocial/app_state.py 10 0 100%
calsocial/cache.py 73 11 85% 65-70, 98, 113, 124, 137, 156-159
calsocial/calendar_system/__init__.py 10 3 70% 32, 41, 48
calsocial/calendar_system/gregorian.py 77 0 100%
calsocial/config_development.py 11 11 0% 4-17
calsocial/config_testing.py 12 0 100%
calsocial/forms.py 198 83 58% 49, 59, 90, 136-146, 153, 161-169, 188-195, 198-206, 209-212, 228-232, 238-244, 252-253, 263-267, 273-277, 317-336, 339-342, 352-354, 362-374, 401-413
calsocial/models.py 372 92 75% 49-51, 103-106, 177, 180-188, 191-200, 203, 242-248, 257-268, 289, 307, 349, 352-359, 378, 392, 404-409, 416, 444, 447, 492-496, 503, 510, 516, 522, 525, 528, 535-537, 545-551, 572, 606-617, 620, 652, 655, 660, 700, 746-748, 762-767, 774-783, 899, 929, 932
calsocial/security.py 15 3 80% 36, 56-58
calsocial/utils.py 42 5 88% 45-48, 52-53
----------------------------------------------------------------------
TOTAL 1104 324 71%
```
All we have to do is converting these ranges into a list of numbers, and compare it with the
result of the previous function:
```python
def get_uncovered_lines(changed_lines):
"""Get the full list of lines that has not been covered by tests
"""
column_widths = []
uncovered_lines = {}
for line in sys.stdin:
line = line.strip()
if line.startswith('---'):
continue
if line.startswith('Name '):
match = re.match(r'^(Name +)(Stmts +)(Miss +)(Cover +)Missing$', line)
assert match
column_widths = [len(col) for col in match.groups()]
continue
name = [
line[sum(column_widths[0:idx]):sum(column_widths[0:idx]) + width].strip()
for idx, width in enumerate(column_widths)][0]
missing = line[sum(column_widths):].strip()
for value in missing.split(', '):
if not value:
continue
try:
number = int(value)
except ValueError:
first, last = value.split('-')
lines = range(int(first), int(last) + 1)
else:
lines = range(number, number + 1)
for lineno in lines:
if name in changed_lines and lineno not in changed_lines[name]:
uncovered_lines.setdefault(name, [])
uncovered_lines[name].append(lineno)
return uncovered_lines
```
At the end we have a dictionary that has filenames as keys, and a list of changed but uncovered
lines.
### Converting back to ranges
To make the final result more readable, lets convert them back to a nice `from_line-to_line`
range list first:
```python
def line_numbers_to_ranges():
"""List the lines that has not been covered
"""
changed_lines = get_changed_lines()
uncovered_lines = get_uncovered_lines(changed_lines)
line_list = []
for filename, lines in uncovered_lines.items():
lines = sorted(lines)
last_value = None
ranges = []
for lineno in lines:
if last_value and last_value + 1 == lineno:
ranges[-1].append(lineno)
else:
ranges.append([lineno])
last_value = lineno
range_list = []
for range_ in ranges:
first = range_.pop(0)
if range_:
range_list.append(f'{first}-{range_[-1]}')
else:
range_list.append(str(first))
line_list.append((filename, ', '.join(range_list)))
return line_list
```
### Printing the result
Now all that is left is to print the result on the screen in a format digestable by a human being:
```python
def tabular_print(uncovered_lines):
"""Print the list of uncovered lines on the screen in a tabbed format
"""
max_filename_len = max(len(data[0]) for data in uncovered_lines)
for filename, lines in uncovered_lines:
print(filename.ljust(max_filename_len + 2) + lines)
```
And we are done.
### Conclusion
This task never seemed hard to accomplish, but somehow I never put enough energy into it to make
it happen. Kudos to Adam Young doing some legwork for me!