Getting git describe into a PyInstaller program
I’ve written about the joy of distributing scripts to less tech-savvy coworkers. Now that some of these simple GUIs have been out in the field for a few weeks, I’m getting a few reports of buggy or odd behavior from the end consumers. As a result, it’s my turn to learn a hard programmer’s lesson: it’s super challenging to fix bigs without a corresponding version to test/repro against. After some putzing and poking online, I learned about git describe. Just the ticket for my versioning needs! And, even better, there’s a great Python wrapper for it - I just have to import git in my Python scripts, and I can extract information from git seamlessly!
…or so I thought.
The trouble, I’ve come to realize, is that PyInstaller and GitPython don’t play well together. Just adding the line import git into a PyInstaller script causes your compiled Python .exe to fail with a pretty cryptic error (“Failed to execute script”, and an OK button). This is a known issue, but I don’t see a clear line of sight to a fix (or even a diagnosis of the problem) in that GitHub issue.
So, yeah - workaround time.
Instead of embedding git describe into the program itself via GitPython, I ended up moving into a distribution script. This is a really simple config wrapper that goes through all the manual steps of generating the final .exe, so I don’t have to remember the build process each time. Now, in addition to all of that build work, the distribution script generates a version.txt file every time it is run. The actual program reads in the flat version.txt file on startup, and uses that string as its version information.
It’s a start, but PyInstaller isn’t smart enough to include this version.txt file in the installer without you telling it to. I’ve found the easiest way to do that is through editing PyInstaller’s .spec files. This is pretty straightforward if you follow their documentation. I ended up following the documentation instructions, and modified the Analysis initilization with something like this:
added_files = [('version.txt', '.')]
a = Analysis(['example_script.py'],
pathex=['C:\\Users\\nash.reilly\\Documents\\GitHub\\example_script'],
binaries=[],
datas=added_files,
hiddenimports=[],
hookspath=[],
runtime_hooks=[],
excludes=[],
win_no_prefer_redirects=False,
win_private_assemblies=False,
cipher=block_cipher,
noarchive=False)
…where all I needed to do was specify added_files = [('version.txt', '.')], and pass that to the Analysis.datas field.
Note, too, that once you’ve edited the .spec file, you’ll also need to pass the .spec file to PyInstaller, rather than your actual script. If you don’t, PyInstaller will just generate a new .spec file with each call - which will overwrite any changes you’ve included locally in your spec file. Here’s a pretty bare-bones version of what my distribution script looks like. See that PyInstaller is using example_script.spec rather than the input Python file.
import os
import PyInstaller.__main__
import git
if __name__ == '__main__':
# Deletes current version.txt file
if os.path.exists('version.txt'):
os.remove('version.txt')
# Writes latest git version info to 'version.txt'
r = git.repo.Repo(search_parent_directories=True)
version_info = r.git.describe('--dirty', '--tags')
with open('version.txt', 'w') as f:
f.write(version_info)
f.write('\n')
f.close()
# Builds production version (no debug console window)
PyInstaller.__main__.run([
'--name=%s' % 'Example Application',
'--noconsole',
'--onefile',
'--noconfirm',
'example_script.spec'
])
Good news - if you’re zipping and distributing the whole output of PyInstaller (i.e. lots of files in your dist directory), you’re done!
However, I chose instead make the PyInstaller exe a single file application, for a few reasons:
- Easier distribution to coworkers
- No confusion about which system icon to click - this script is going to people who know nothing about Python or programming, so ease of use is paramount.
This needs a little bonus effort to work right. The problem, as explained by this StackOverflow post, is that version.txt lives in different places in the dev and deployed environments when you use the --onefile option. In your dev environment, this is the same directory as the build script. (Hence this: “.”) When you hand the bundled script off to other people, PyInstaller unpacks that version.txt folder into the system’s temp data folder. You can get access there by calling sys._MEIPASS. The helper function resource_path() is a quick sanity check that modifies the base path of version.txt based on some really simple exception logic, which allows you to check both locations gracefully for your versioning info:
def resource_path(relative_path):
""" Get absolute path to resource, works for dev and for PyInstaller """
try:
# PyInstaller creates a temp folder and stores path in _MEIPASS
base_path = sys._MEIPASS
except Exception:
base_path = os.path.abspath(".")
return os.path.join(base_path, relative_path)
# Git Repo information
version_info = ''
version_file_path = resource_path('version.txt')
with open(version_file_path) as f:
version_info = f.readlines()[0].strip()
Note well: every config file that gets bundled with the .exe needs to be read in to your program thru the resource_path() function. Otherwise, your program will crash, because it will look in the local directory for that file, and not find it!