Bundling data files with PyInstaller (--onefile)
I'm trying to build a one-file EXE with PyInstaller which is to include an image and an icon. I cannot for the life of me get it to work with --onefile.
If I do --onedir it works all works very well.
When I use --onefile, it can't find the referenced additional files (when running the compiled EXE). It finds the DLLs开发者_StackOverflow中文版 and everything else fine, just not the two images.
I've looked in the temp-dir generated when running the EXE (\Temp\_MEI95642\ for example) and the files are indeed in there. When I drop the EXE in that temp-directory it finds them. Very perplexing.
This is what I've added to the .spec file
a.datas += [('images/icon.ico', 'D:\\[workspace]\\App\\src\\images\\icon.ico', 'DATA'),
('images/loaderani.gif','D:\\[workspace]\\App\\src\\images\\loaderani.gif','DATA')]
I should add that I have tried not putting them in subfolders as well, didn't make a difference.
Edit: Marked newer answer as correct due to PyInstaller update.
Newer versions of PyInstaller do not set the env variable anymore, so Shish's excellent answer will not work. Now the path gets set as sys._MEIPASS:
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)
pyinstaller unpacks your data into a temporary folder, and stores this directory path in the _MEIPASS2 environment variable. To get the _MEIPASS2 dir in packed-mode and use the local directory in unpacked (development) mode, I use this:
def resource_path(relative):
return os.path.join(
os.environ.get(
"_MEIPASS2",
os.path.abspath(".")
),
relative
)
Output:
# in development
>>> resource_path("app_icon.ico")
"/home/shish/src/my_app/app_icon.ico"
# in production
>>> resource_path("app_icon.ico")
"/tmp/_MEI34121/app_icon.ico"
All of the other answers use the current working directory in the case where the application is not PyInstalled (i.e. sys._MEIPASS is not set). That is wrong, as it prevents you from running your application from a directory other than the one where your script is.
A better solution:
import sys
import os
def resource_path(relative_path):
""" Get absolute path to resource, works for dev and for PyInstaller """
base_path = getattr(sys, '_MEIPASS', os.path.dirname(os.path.abspath(__file__)))
return os.path.join(base_path, relative_path)
Perhaps i missed a step or did something wrong but the methods which are above, didn't bundle data files with PyInstaller into one exe file. Let me share the steps what i have done.
step:Write one of the above methods into your py file with importing sys and os modules. I tried both of them. The last one is:
def resource_path(relative_path): """ Get absolute path to resource, works for dev and for PyInstaller """ base_path = getattr(sys, '_MEIPASS', os.path.dirname(os.path.abspath(__file__))) return os.path.join(base_path, relative_path)step: Write, pyi-makespec file.py, to the console, to create a file.spec file.
step: Open, file.spec with Notepad++ to add the data files like below:
a = Analysis(['C:\\Users\\TCK\\Desktop\\Projeler\\Converter-GUI.py'], pathex=['C:\\Users\\TCK\\Desktop\\Projeler'], binaries=[], datas=[], hiddenimports=[], hookspath=[], runtime_hooks=[], excludes=[], win_no_prefer_redirects=False, win_private_assemblies=False, cipher=block_cipher) #Add the file like the below example a.datas += [('Converter-GUI.ico', 'C:\\Users\\TCK\\Desktop\\Projeler\\Converter-GUI.ico', 'DATA')] pyz = PYZ(a.pure, a.zipped_data, cipher=block_cipher) exe = EXE(pyz, a.scripts, exclude_binaries=True, name='Converter-GUI', debug=False, strip=False, upx=True, #Turn the console option False if you don't want to see the console while executing the program. console=False, #Add an icon to the program. icon='C:\\Users\\TCK\\Desktop\\Projeler\\Converter-GUI.ico') coll = COLLECT(exe, a.binaries, a.zipfiles, a.datas, strip=False, upx=True, name='Converter-GUI')step: I followed the above steps, then saved the spec file. At last opened the console and write, pyinstaller file.spec (in my case, file=Converter-GUI).
Conclusion: There's still more than one file in the dist folder.
Note: I'm using Python 3.5.
EDIT: Finally it works with Jonathan Reinhart's method.
step: Add the below codes to your python file with importing sys and os.
def resource_path(relative_path): """ Get absolute path to resource, works for dev and for PyInstaller """ base_path = getattr(sys, '_MEIPASS', os.path.dirname(os.path.abspath(__file__))) return os.path.join(base_path, relative_path)step: Call the above function with adding the path of your file:
image_path = resource_path("Converter-GUI.ico")step: Write the above variable that call the function to where your codes need the path. In my case it's:
self.window.iconbitmap(image_path)step: Open the console in the same directory of your python file, write the codes like below:
pyinstaller --onefile your_file.py- step: Open the .spec file of the python file and append the a.datas array and add the icon to the exe class, which was given above before the edit in 3'rd step.
step: Save and exit the path file. Go to your folder which include the spec and py file. Open again the console window and type the below command:
pyinstaller your_file.spec
After the 6. step your one file is ready to use.
I have been dealing with this issue for a long(well, very long) time. I've searched almost every source but things were not getting in a pattern in my head.
Finally, I think I have figured out exact steps to follow, I wanted to share.
Note that, my answer uses information on the answers of others on this question.
How to create a standalone executable of a python project.
Assume, we have a project_folder and the file tree is as follows:
project_folder/ main.py xxx.py # another module yyy.py # another module sound/ # directory containing the sound files img/ # directory containing the image files venv/ # if using a venv
First of all, let's say you have defined your paths to sound/ and img/ folders into variables sound_dir and img_dir as follows:
img_dir = os.path.join(os.path.dirname(__file__), "img")
sound_dir = os.path.join(os.path.dirname(__file__), "sound")
You have to change them, as follows:
img_dir = resource_path("img")
sound_dir = resource_path("sound")
Where, resource_path() is defined in the top of your script as:
def resource_path(relative_path):
""" Get absolute path to resource, works for dev and for PyInstaller """
base_path = getattr(sys, '_MEIPASS', os.path.dirname(os.path.abspath(__file__)))
return os.path.join(base_path, relative_path)
Activate virtual env if using a venv,
Install pyinstaller if you didn't yet, by: pip3 install pyinstaller.
Run: pyi-makespec --onefile main.py to create the spec file for the compile and build process.
This will change file hierarchy to:
project_folder/ main.py xxx.py # modules xxx.py # modules sound/ # directory containing the sound files img/ # directory containing the image files venv/ # if using a venv main.spec
Open(with an edior) main.spec:
At top of it, insert:
added_files = [
("sound", "sound"),
("img", "img")
]
Then, change the line of datas=[], to datas=added_files,
For the details of the operations done on main.spec see here.
Run pyinstaller --onefile main.spec
And that is all, you can run main in project_folder/dist from anywhere, without having anything else in its folder. You can distribute only that main file. It is now, a true standalone.
Using the excellent answer from Max and This post about adding extra data files like images or sound & my own research/testing, I've figured out what I believe is the easiest way to add such files.
If you would like to see a live example, my repository is here on GitHub.
Note: this is for compiling using the --onefile or -F command with pyinstaller.
My environment is as follows.
- Python 3.3.7
- Tkinter 8.6 (check version)
- Pyinstaller 3.6
Solving the problem in 2 steps
To solve the issue we need to specifically tell Pyinstaller that we have extra files that need to be "bundled" with the application.
We also need to be using a 'relative' path, so the application can run properly when it's running as a Python Script or a Frozen EXE.
With that being said we need a function that allows us to have relative paths. Using the function that Max Posted we can easily solve the relative pathing.
def img_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)
We would use the above function like this so the application icon shows up when the app is running as either a Script OR Frozen EXE.
icon_path = img_resource_path("app/img/app_icon.ico")
root.wm_iconbitmap(icon_path)
The next step is that we need to instruct Pyinstaller on where to find the extra files when it's compiling so that when the application is run, they get created in the temp directory.
We can solve this issue two ways as shown in the documentation, but I personally prefer managing my own .spec file so that's how we're going to do it.
First, you must already have a .spec file. In my case, I was able to create what I needed by running pyinstaller with extra args, you can find extra args here. Because of this, my spec file may look a little different than yours but I'm posting all of it for reference after I explain the important bits.
added_files is essentially a List containing Tuple's, in my case I'm only wanting to add a SINGLE image, but you can add multiple ico's, png's or jpg's using ('app/img/*.ico', 'app/img') You may also create another tuple like soadded_files = [ (), (), ()] to have multiple imports
The first part of the tuple defines what file or what type of file's you would like to add as well as where to find them. Think of this as CTRL+C
The second part of the tuple tells Pyinstaller, to make the path 'app/img/' and place the files in that directory RELATIVE to whatever temp directory gets created when you run the .exe. Think of this as CTRL+V
Under a = Analysis([main..., I've set datas=added_files, originally it used to be datas=[] but we want out the list of imports to be, well, imported so we pass in our custom imports.
You don't need to do this unless you want a specific icon for the EXE, at the bottom of the spec file I'm telling Pyinstaller to set my application icon for the exe with the option icon='app\\img\\app_icon.ico'.
added_files = [
('app/img/app_icon.ico','app/img/')
]
a = Analysis(['main.py'],
pathex=['D:\\Github Repos\\Processes-Killer\\Process Killer'],
binaries=[],
datas=added_files,
hiddenimports=[],
hookspath=[],
runtime_hooks=[],
excludes=[],
win_no_prefer_redirects=False,
win_private_assemblies=False,
cipher=block_cipher,
noarchive=False)
pyz = PYZ(a.pure, a.zipped_data,
cipher=block_cipher)
exe = EXE(pyz,
a.scripts,
a.binaries,
a.zipfiles,
a.datas,
[],
name='Process Killer',
debug=False,
bootloader_ignore_signals=False,
strip=False,
upx=True,
upx_exclude=[],
runtime_tmpdir=None,
console=True , uac_admin=True, icon='app\\img\\app_icon.ico')
Compiling to EXE
I'm very lazy; I don't like typing things more than I have to. I've created a .bat file that I can just click. You don't have to do this, this code will run in a command prompt shell just fine without it.
Since the .spec file contains all of our compiling settings & args (aka options) we just have to give that .spec file to Pyinstaller.
pyinstaller.exe "Process Killer.spec"
Instead for rewriting all my path code as suggested, I changed the working directory:
if getattr(sys, 'frozen', False):
os.chdir(sys._MEIPASS)
Just add those two lines at the beginning of your code, you can leave the rest as is.
Slight modification to the accepted answer.
def resource_path(relative_path):
""" Get absolute path to resource, works for dev and for PyInstaller """
if hasattr(sys, '_MEIPASS'):
return os.path.join(sys._MEIPASS, relative_path)
return os.path.join(os.path.abspath("."), relative_path)
Another solution is to make a runtime hook, which will copy(or move) your data (files/folders) to the directory at which the executable is stored. The hook is a simple python file that can almost do anything, just before the execution of your app. In order to set it, you should use the --runtime-hook=my_hook.py option of pyinstaller. So, in case your data is an images folder, you should run the command:
pyinstaller.py --onefile -F --add-data=images;images --runtime-hook=cp_images_hook.py main.py
The cp_images_hook.py could be something like this:
import sys
import os
import shutil
path = getattr(sys, '_MEIPASS', os.getcwd())
full_path = path+"\\images"
try:
shutil.move(full_path, ".\\images")
except:
print("Cannot create 'images' folder. Already exists.")
Before every execution the images folder is moved to the current directory (from the _MEIPASS folder), so the executable will always have access to it. In that way there is no need to modify your project's code.
Second Solution
You can take advantage of the runtime-hook mechanism and change the current directory, which is not a good practice according to some developers, but it works fine.
The hook code can be found below:
import sys
import os
path = getattr(sys, '_MEIPASS', os.getcwd())
os.chdir(path)
The most common complaint/question I've seen wrt PyInstaller is "my code can't find a data file which I definitely included in the bundle, where is it?", and it isn't easy to see what/where your code is searching because the extracted code is in a temp location and is removed when it exits. Add this bit of code to see what's included in your onefile and where it is, using @Jonathon Reinhart's resource_path()
for root, dirs, files in os.walk(resource_path("")):
print(root)
for file in files:
print( " ",file)
If you are still trying to put files relative to your executable instead of in the temp directory, you need to copy it yourself. This is how I ended up getting it done.
https://stackoverflow.com/a/59415662/999943
You add a step in the spec file that does a filesystem copy to the DISTPATH variable.
Hope that helps.
i use this based on max solution
def getPath(filename):
import os
import sys
from os import chdir
from os.path import join
from os.path import dirname
from os import environ
if hasattr(sys, '_MEIPASS'):
# PyInstaller >= 1.6
chdir(sys._MEIPASS)
filename = join(sys._MEIPASS, filename)
elif '_MEIPASS2' in environ:
# PyInstaller < 1.6 (tested on 1.5 only)
chdir(environ['_MEIPASS2'])
filename = join(environ['_MEIPASS2'], filename)
else:
chdir(dirname(sys.argv[0]))
filename = join(dirname(sys.argv[0]), filename)
return filename
FINALLY ! based on solution from Luca
I was able to finaly execute files inside included directory in --onefile pyinstaller.
Just use the method he posted and you can even enter a whole directory name to files and it still works.
Second thing you will need is actually add the directory/files you want inside pyinstaller. Just use args like:
--add-data "yourDir/And/OrFile;yourDir"
I found the existing answers confusing, and took a long time to work out where the problem is. Here's a compilation of everything I found.
When I run my app, I get an error Failed to execute script foo (if foo.py is the main file). To troubleshoot this, don't run PyInstaller with --noconsole (or edit main.spec to change console=False => console=True). With this, run the executable from a command-line, and you'll see the failure.
The first thing to check is that it's packaging up your extra files correctly. You should add tuples like ('x', 'x') if you want the folder x to be included.
After it crashes, don't click OK. If you're on Windows, you can use Search Everything. Look for one of your files (eg. sword.png). You should find the temporary path where it unpacked the files (eg. C:\Users\ashes999\AppData\Local\Temp\_MEI157682\images\sword.png). You can browse this directory and make sure it included everything. If you can't find it this way, look for something like main.exe.manifest (Windows) or python35.dll (if you're using Python 3.5).
If the installer includes everything, the next likely problem is file I/O: your Python code is looking in the executable's directory, instead of the temp directory, for files.
To fix that, any of the answers on this question work. Personally, I found a mixture of them all to work: change directory conditionally first thing in your main entry-point file, and everything else works as-is:
if hasattr(sys, '_MEIPASS'):
os.chdir(sys._MEIPASS)
For those of whom are still looking for a more recent answer, here you go:
In the documentation, there's a section on accessing added data files.
Here is the short and sweet of it.
You'll want to import pkgutil and locate which folder you added the datafile to; i.e. the second string in the tuple which was added to the spec file:
datas = [("path/to/mypackage/data_file.txt", "path/to/mypackage")]
Knowing where you added the data file can then be used for reading it in as binary data, and decoding it as you wish. Take this example:
File structure:
mypackage
__init__.py # This is a MUST in order for the package to be registered
data_file.txt # The data file you've added
data_file.txt
Hello world!
main.py
import pkgutil
file = pkgutil.get_data("mypackage", "data_file.txt")
contents = file.decode("utf-8")
print(contents) # Hello world!
References:
pkgutil- Builtin library__init__.pyand packages
加载中,请稍侯......
精彩评论