feat(scripts): Generate summary of shared-vs-platform-specific code.

This is a little experiment to get a sense of whether our "shared code"
strategy is actually paying off.  It generates a simple lines-of-code
analysis of how much of our code is shared for different platforms,
versus platform-specific binding code.
This commit is contained in:
Ryan Kelly 2019-04-02 13:43:32 +11:00
Родитель 622b0ddb93
Коммит 5c8d52574a
Не найден ключ, соответствующий данной подписи
Идентификатор ключа GPG: FB70C973A037D258
1 изменённых файлов: 122 добавлений и 0 удалений

122
tools/loc_summary.py Normal file
Просмотреть файл

@ -0,0 +1,122 @@
#
# This is a simple script to generate some summary lines-of-code metrics.
# Use it like this:
#
# $> python locsummary.py [./path/to/code/directory]
#
# It shells out to tokei [1] to do the actual counting, which you must have
# installed with the `json` feature. It then massages the tokei output to
# give us a rough guesstimate at:
#
# * How much code is shared code, that can be used on multiple platforms.
# * How much of it is android-specific code.
# * How much of it is ios-specific code.
#
# Since we aim to reduce total-cost-of-ownership of our storage and sync
# infrastructure by re-using code across platforms, we should hope that
# a high percentage of the code we've written is shared rather than
# platform-specific! The analysis is far from an exact science, but it
# provides a nice gut-check for our code re-use story.
#
# [1] https://github.com/XAMPPRocky/tokei
#
import os.path
import argparse
import subprocess
import json
ROOT_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
DEFAULT_PATH = os.path.join(ROOT_DIR, 'components')
# For each type of file in our repo, is it:
# * shared code across platforms?
# * specific to android?
# * specific to ios?
# * some sort of meta file that we shouldn't count at all?
FILETYPE_TO_SUMMARYTYPE = {
'CHeader': 'ios',
'Java': 'android',
'Json': None,
'Kotlin': 'android',
'Markdown': None,
'Prolog': 'android', # Our .pro files are actually proguard rules, not prolog...
'Protobuf': 'shared',
'Rust': 'shared',
'Sql': 'shared',
'Swift': 'ios',
'Toml': None,
'XcodeConfig': 'ios',
'Xml': None,
}
def get_loc_summary(path):
path = os.path.abspath(path)
p = subprocess.run([
'tokei',
'--output', 'json',
'--exclude', 'examples',
path,
], stdout=subprocess.PIPE, universal_newlines=True)
p.check_returncode()
summary = {
'shared': 0,
'android': 0,
'ios': 0,
'all': 0,
}
lineOfCode = json.loads(p.stdout)['inner']
for fileType in lineOfCode:
summaryType = FILETYPE_TO_SUMMARYTYPE[fileType]
if summaryType is not None:
summary[summaryType] += lineOfCode[fileType]['code']
summary['all'] += lineOfCode[fileType]['code']
if path.startswith(ROOT_DIR):
summary['path'] = os.path.join('.', path[len(ROOT_DIR) + 1:])
else:
summary['path'] = path
return summary
def print_loc_summaries(paths):
summaries = [get_loc_summary(path) for path in paths]
headers = ['Path', 'Shared', 'Android', 'iOS', 'Total', 'Shared %']
nameWidth = max(
len(headers[0]),
max(len(summary['path']) for summary in summaries)
)
numWidth = max(
max(len(h) for h in headers[1:]),
max(len(str(summary['all'])) for summary in summaries)
)
totalWidth = (nameWidth + 5) + (numWidth + 3) * 5 - 1
print("-" * totalWidth)
print(
f"| {headers[0]:<{nameWidth}} | "
f"{headers[1]:>{numWidth}} | "
f"{headers[2]:>{numWidth}} | "
f"{headers[3]:>{numWidth}} | "
f"{headers[4]:>{numWidth}} | "
f"{headers[5]:>{numWidth}} |"
)
print("-" * totalWidth)
for summary in summaries:
print(
f"| {summary['path']:<{nameWidth}} | "
f"{summary['shared']:>{numWidth}} | "
f"{summary['android']:>{numWidth}} | "
f"{summary['ios']:>{numWidth}} | "
f"{summary['all']:>{numWidth}} | "
f"{(summary['shared'] / summary['all']):>{numWidth}.2%} |"
)
print("-" * totalWidth)
if __name__ == "__main__":
parser = argparse.ArgumentParser(description="summarize lines-of-code statistics")
parser.add_argument('paths', type=str, nargs='*', default=[DEFAULT_PATH])
args = parser.parse_args()
print_loc_summaries(args.paths)