Skip to content

Latest commit

 

History

History
235 lines (186 loc) · 7.78 KB

readme_building_executables.md

File metadata and controls

235 lines (186 loc) · 7.78 KB
Raw

Building Executables for macOS, Linux, and Windows - Running in Browser Option

This readme outlines the steps to create executables for macOS, Linux, and Windows using Python, Poetry, and PyInstaller. The process is similar across platforms with some platform-specific adjustments.

Initial Setup

1. Clean Previous Builds (if necessary)

Before starting, it's good to clean up any previous builds:

rm -rf dist build *.spec
sudo rm -rf /mnt/c/MyFiles/050-FDSN/seed-vault/.venv

2. Install Python, Poetry, and PyInstaller

For each platform (Linux, macOS, Windows):

  • Install Python from python.org.
  • Install Poetry by running the following command: 
    (Invoke-WebRequest -Uri https://install.python-poetry.org -UseBasicParsing).Content | python -
  • Install PyInstaller using Poetry or pip: 
    poetry add --dev pyinstaller

3. Install Project Dependencies with Poetry

Once Poetry is installed, install all dependencies for your project:

poetry install
poetry shell
poetry show --with dev

Streamlit Adjustments

[https://github.com/jvcss/PyInstallerStreamlit]

Modify Streamlit to Run in PyInstaller

Add Content to run_app.py

Create a file called run_app.py with the following content:

from streamlit.web import cli

if __name__ == '__main__':
    cli._main_run_clExplicit('seed_vault/ui/main.py', is_hello=False)

Modify Streamlit's CLI

Navigate to the Streamlit path in your virtual environment:

.env\Lib\site-packages\streamlit\web\cli.py

Add the following function to cli.py:

def _main_run_clExplicit(file, is_hello, args=[], flag_options={}):
    bootstrap.run(file, is_hello, args, flag_options)

Create a Hook to Collect Streamlit Metadata

Create a new hook file hook-streamlit.py under the ./hooks directory:

from PyInstaller.utils.hooks import copy_metadata

datas = copy_metadata('streamlit')

Compile the App with PyInstaller

For Linux

Use the following command to compile the app on Linux:

pyinstaller --name seed-vault \
            --onefile \
            --additional-hooks-dir=./hooks \
            --collect-all streamlit \
            --collect-all folium \
            --collect-all obspy \
            --collect-all streamlit-folium \
            --specpath . \
            --clean \
            run_app.py

For Windows

For Windows, use PowerShell and the following command:

pyinstaller --name seed-vault `
            --onefile `
            --additional-hooks-dir=.\hooks `
            --collect-all streamlit `
            --collect-all folium `
            --collect-all obspy `
            --collect-all streamlit-folium `
            --specpath . `
            --clean `
            run_app.py

For macOS

For macOS, the command is similar to Linux:

pyinstaller --name seed-vault \
            --onefile \
            --additional-hooks-dir=./hooks \
            --collect-all streamlit \
            --collect-all folium \
            --collect-all obspy \
            --collect-all streamlit-folium \
            --specpath . \
            --clean \
            run_app.py

Streamlit Configuration

Create the following configuration file for Streamlit to specify server options.

.streamlit/config.toml:

[global]
developmentMode = false

[server]
port = 8502

Add this file either to the project root or the dist output folder after building.

Copy Configuration and Files

After building, copy configuration and source files to the dist folder:

xcopy /s /e ".\.streamlit" "dist\.streamlit"
Copy-Item -Path "seed_vault" -Destination "dist\seed_vault" -Recurse

Modify the Spec File for Each Platform

Linux Spec File

For Linux, the datas and hiddenimports sections in the .spec file should look like this:

datas = datas + [
    (".venv/lib/python3.12/site-packages/altair/vegalite/v5/schema/vega-lite-schema.json", "./altair/vegalite/v5/schema/"),
    (".venv/lib/python3.12/site-packages/streamlit/static", "./streamlit/static"),
    (".venv/lib/python3.12/site-packages/streamlit/runtime", "./streamlit/runtime"),
    (".venv/lib/python3.12/site-packages/streamlit_folium", "streamlit_folium"),
    ('.venv/lib/python3.12/site-packages/obspy/RELEASE-VERSION', 'obspy/'),
    ("seed_vault", "seed_vault"), 
    ("data", "data"), 
    (".streamlit/config.toml", ".streamlit/config.toml"),
]

hiddenimports = hiddenimports + ['tqdm', 'streamlit-folium', 'pandas', 'requests', 'click', 'tabulate', 'numpy', 'matplotlib', 'pydantic']

Windows Spec File

For Windows, adjust the paths to Windows-style:

datas = datas + [
    (".venv\Lib\site-packages\altair\vegalite\v5\schema\vega-lite-schema.json", "./altair/vegalite/v5/schema/"),
    (".venv\Lib\site-packages\streamlit\static", "./streamlit/static"),
    (".venv\Lib\site-packages\streamlit\runtime", "./streamlit/runtime"),
    (".venv\Lib\site-packages\streamlit_folium", "streamlit_folium"),
    ('.venv\Lib\site-packages\obspy\RELEASE-VERSION', 'obspy/'),
    ("seed_vault", "seed_vault"), 
    ("data", "data"), 
    (".streamlit\config.toml", ".streamlit/config.toml"),
]

hiddenimports = hiddenimports + ['tqdm', 'streamlit-folium', 'pandas', 'requests', 'click', 'tabulate', 'numpy', 'matplotlib', 'pydantic']

macOS Spec File

For macOS, use the same style as Linux, with paths similar to Linux paths:

datas = datas + [
    (".venv/lib/python3.12/site-packages/altair/vegalite/v5/schema/vega-lite-schema.json", "./altair/vegalite/v5/schema/"),
    (".venv/lib/python3.12/site-packages/streamlit/static", "./streamlit/static"),
    (".venv/lib/python3.12/site-packages/streamlit/runtime", "./streamlit/runtime"),
    (".venv/lib/python3.12/site-packages/streamlit_folium", "streamlit_folium"),
    ('.venv/lib/python3.12/site-packages/obspy/RELEASE-VERSION', 'obspy/'),
    ("seed_vault", "seed_vault"), 
    ("data", "data"), 
    (".streamlit/config.toml", ".streamlit/config.toml"),
]

hiddenimports = hiddenimports + ['tqdm', 'streamlit-folium', 'pandas', 'requests', 'click', 'tabulate', 'numpy', 'matplotlib', 'pydantic']

Build the Executable

Finally, to build the executable on any platform, run:

pyinstaller run_app.spec --clean

Conclusion

By following these steps, you can build standalone executables for macOS, Linux, and Windows using Poetry and PyInstaller. Each platform requires slight adjustments, such as path formatting, but the overall process remains consistent across environments.

Building as Desktop app

How to build

@stlite/desktop can be used to build a desktop app with Steamlit. Follow the below step:

  • Create a package.json. There is already one created in the project (see https://github.com/whitphx/stlite/blob/main/packages/desktop/README.md)
  • Make sure all required project files are available in package.json
  • Make sure requirements.txt is available for lib dependencies
  • npm install -> this will install required node_modules
  • npm run dump -> builds the app
  • npm run serve -> serves the app
  • npm run dist -> creates executable

Main Challenge - Lib dependecies

stlite only accept libraries that have pure wheels, i.e., they are built for webassembly. pyodide (https://pyodide.org/en/stable/usage/faq.html#why-can-t-micropip-find-a-pure-python-wheel-for-a-package) seems to be reponsible to bundle the app. This library already comes with a list of famous libs such as pandas. But it does not support less famous/widely used libs such as obspy.

Packagin a lib in to a webassembly wheel seems to be quite complicated and not worthy of much try. And this is the main blocker of this approach.

NOTE: packages with pure wheel will have *py3-none-any.whl in their naming.