PyInstaller打包含Plotly Express程序时Numpy依赖缺失问题

问题背景 #

我开发了一个基于tkinter、pandas和plotly.express的程序，在PyCharm中运行完全正常，导入代码如下：

import tkinter as tk
import tkinter.filedialog as fd
import plotly.express as px
import pandas

但用PyInstaller为Windows系统打包生成EXE文件后，运行时抛出以下错误：

Traceback (most recent call last):
  File "[PATH]\main.py", line 4, in <module>
    import plotly.express as px
  File "PyInstaller\loader\pyimod02_importers.py", line 384, in exec_module
  File "plotly\express\__init__.py", line 10, in <module>
ImportError: Plotly Express requires numpy to be installed. You can install numpy using pip with:

$ pip install numpy

Or install Plotly Express and its dependencies directly with:

$ pip install "plotly[express]"

You can also use Plotly Graph Objects to create a large number of charts without installing
numpy. See examples here: https://plotly.com/python/graph-objects/

[PYI-11036:ERROR] Failed to execute script 'main' due to unhandled exception!

尝试过的无效操作：

• 重新安装numpy和plotly-express（原本已正常安装）
• 在main.py中直接导入numpy，触发you should not try to import numpy from its source directory错误
• 将plotly目录添加到main.spec的datas中，打包后问题依旧
• 在main.spec的hiddenimports中添加['numpy', 'numpy.core', 'numpy.lib.format']，打包后warn-main.txt仍提示numpy._core.XXXX类模块缺失

解决方案 #

1. 用干净虚拟环境重装依赖 #

PyInstaller打包易受全局环境混乱影响，建议创建全新虚拟环境：

# 创建并激活虚拟环境
python -m venv venv
venv\Scripts\activate

# 重装所有依赖，明确安装完整Plotly Express依赖
pip install pandas plotly[express] pyinstaller

2. 修正Spec文件配置 #

更新main.spec，重点优化以下几点：

• 补充更多numpy核心子模块到hiddenimports
• 禁用UPX压缩（部分场景下UPX会破坏numpy二进制文件）
• 无需手动添加plotly到datas，依赖PyInstaller内置钩子处理

修改后的main.spec示例：

# -*- mode: python ; coding: utf-8 -*- 

a = Analysis(
    ['main.py'],
    pathex=[],
    binaries=[],
    datas=[],
    hiddenimports=[
        'numpy',
        'numpy.core',
        'numpy.core.multiarray',
        'numpy.core.numeric',
        'numpy.lib.format',
        'numpy.random',
        'plotly',
        'plotly.express',
        'plotly.io._html',
        'plotly.io._json'
    ],
    hookspath=[],
    hooksconfig={},
    runtime_hooks=[],
    excludes=[],
    noarchive=False,
    optimize=0,
)
pyz = PYZ(a.pure)

exe = EXE(
    pyz,
    a.scripts,
    [],
    exclude_binaries=True,
    name='main',
    debug=False,
    bootloader_ignore_signals=False,
    strip=False,
    upx=False,  # 禁用UPX避免损坏numpy二进制
    console=True,
    disable_windowed_traceback=False,
    argv_emulation=False,
    target_arch=None,
    codesign_identity=None,
    entitlements_file=None,
)
coll = COLLECT(
    exe,
    a.binaries,
    a.datas,
    strip=False,
    upx=False,  # 同步禁用UPX
    upx_exclude=[],
    name='main',
)

3. 提前在代码中导入numpy #

在main.py最顶部预导入numpy，强制PyInstaller识别该依赖：

import numpy as np  # 提前导入numpy，避免打包时被遗漏
import tkinter as tk
import tkinter.filedialog as fd
import plotly.express as px
import pandas

注意确保是从虚拟环境的安装目录导入numpy，而非源码文件夹。

4. 更新PyInstaller到最新版 #

旧版本PyInstaller对numpy和plotly的兼容性较差，执行升级：

pip install --upgrade pyinstaller

内容的提问来源于stack exchange，提问作者LoneRetriever