I have coded a tiny python program using PyQt4. Now, I want to use cx_Freeze to create a standalone application. Everything works fine - cx_Freeze includes automatically all necessary modules; the resulting exe works.
The only problem is that cx_Freeze packs plenty of unneeded modules into the standalone. Even though I only use QtCore and QtGui, also modules like sqlite3, QtNetwork, or QtScript are included. Surprisingly, I find also PyQt5 dlls in the resulting folder. It seems to me as if cx_Freeze uses all PyQt packages that I have installed. The result is a 200Mb program - albeit I only wrote a tiny script.
How can I prevent this behaviour?
I use the following setup.py:
import sys
from cx_Freeze import setup, Executable
setup(
name="MyProgram",
version="0.1",
description="MyDescription",
executables=[Executable("MyProgram.py", base = "Win32GUI")],
)
I tried explicitely excluding some packages (although it is quite messy to exclude all unused Qt modules) adding this code:
build_exe_options = {"excludes": ["tkinter", "PyQt4.sqlite3",
"PyQt4.QtOpenGL4", "PyQt4.QtSql"]}
but the upper modules were still used. I also tried
build_exe_options = {"excludes": ["tkinter", "PyQt4.sqlite3",
"QtOpenGL4", "QtSql"]}
with the same result.
In addition to the nedless Qt packages I find also unneded folders with names like "imageformats", "tcl", and "tk". How can I include only needed files in order to keep the standalone folder and installer as small as possible?
I googled this problem for hours but only found this thread which did not help me.
I am running python 3.4.2 amd64 on windows 8.
I am happy about every solution that gives me the desired result "standalone" with a reasonable size. I tried also pyqtdeploy but ran into the error: Unknown module(s) in QT (but this is a different question).
I am using two modules. One is the GUI class created by uic, "MyProgramGUIPreset". In this file there are the following import commands:
from PyQt4 import QtCore, QtGui
from matplotlibwidget import MatplotlibWidget
In the main module I do the following imports:
import MyProgramGUIPreset
import numpy as np
from PyQt4.QtGui import QApplication, QMainWindow, QMessageBox
import sys
from math import *
Maybe this helps to figuring out where the issue is.
The reason for the not working "excludes" command was that I forgot to include the build options into the setup. After adding the respective line into the code excluding works:
from cx_Freeze import setup, Executable
import sys
# exclude unneeded packages. More could be added. Has to be changed for
# other programs.
build_exe_options = {"excludes": ["tkinter", "PyQt4.QtSql", "sqlite3",
"scipy.lib.lapack.flapack",
"PyQt4.QtNetwork",
"PyQt4.QtScript",
"numpy.core._dotblas",
"PyQt5"],
"optimize": 2}
# Information about the program and build command. Has to be adjusted for
# other programs
setup(
name="MyProgram", # Name of the program
version="0.1", # Version number
description="MyDescription", # Description
options = {"build_exe": build_exe_options}, # <-- the missing line
executables=[Executable("MyProgram.py", # Executable python file
base = ("Win32GUI" if sys.platform == "win32"
else None))],
)
This decreased the program size from 230MB to 120MB. Nevertheless, I did not find a nice way of excluding all unneeded packages. By trial and error (deleting the biggest files in the build folder test-wise) I figured out which classes I can exclude.
I tried whether the matplotlib backends cause the problem and finally figured out that this is not the case. Nontheless, if anybody needs code to exclude all modules of a certain name scheme in a particular folder except some special ones, he may adjust the following to his needs:
mplBackendsPath = os.path.join(os.path.split(sys.executable)[0],
"Lib/site-packages/matplotlib/backends/backend_*")
fileList = glob.glob(mplBackendsPath)
moduleList = []
for mod in fileList:
modules = os.path.splitext(os.path.basename(mod))[0]
if not module == "backend_qt4agg":
moduleList.append("matplotlib.backends." + modules)
build_exe_options = {"excludes": ["tkinter"] + moduleList, "optimize": 2}
I would be happy about more elegant solutions. Further ideas are still welcome. Nevertheless, I regard the problem as solved for me.