如何用pip将Python包下载至公开GCS桶并从桶安装及报错解决

Hey there! Let's tackle your problem head-on. First, the error you're seeing (no such option: --download) happens because pip doesn't have a --download flag for the download command. Let's walk through how to correctly download packages to your GCS bucket and install them directly from it.

1. How to Download Python Packages to a Public GCS Bucket #

Pip can't download packages directly to cloud storage like GCS—you'll need to first download them to your local machine, then upload the files to your bucket. Here's how to do it:

Step 1: Download packages to a local directory #

Use the -d (or --dest) flag to specify a local folder where pip will save the package files:

# Download pandas to a temporary local directory
pip3 download pandas -d ./temp_packages

If you need packages for a specific platform or want source distributions instead of wheels, add extra flags. For example, to download source packages:

pip3 download pandas --no-binary=:all: -d ./temp_packages

Step 2: Upload local packages to GCS #

Once you have the files locally, use the Google Cloud CLI (gcloud) to upload them to your bucket. Make sure you've already installed and authenticated the CLI (gcloud auth login):

# Upload all package files to your GCS bucket
gcloud storage cp ./temp_packages/* gs://public_bucket/

If you have nested directories (e.g., for dependencies), add the -r flag to upload recursively:

gcloud storage cp -r ./temp_packages/* gs://public_bucket/

Alternatively, you can use gsutil (the older GCS tool) if you prefer:

gsutil cp ./temp_packages/* gs://public_bucket/

2. How to Install Packages Directly from a Public GCS Bucket #

Since your GCS bucket is publicly accessible, you have two easy ways to install packages via pip:

Option 1: Install directly from the package file URL #

Each package file in your bucket has a public URL in the format:
https://storage.googleapis.com/[bucket-name]/[package-file-name]

For example, if you uploaded pandas-2.1.4-cp311-none-macosx_10_9_x86_64.whl, install it with:

pip3 install https://storage.googleapis.com/public_bucket/pandas-2.1.4-cp311-none-macosx_10_9_x86_64.whl

Option 2: Set up a simple PyPI-style index (for multiple packages) #

If you want to install packages by name (like pip install pandas) instead of using direct URLs, create a PEP 503-compliant index in your GCS bucket:

1. Create a simple directory in your bucket (this is the standard structure for PyPI indexes).
2. Upload all your package files to the simple directory.
3. Generate an index.html file for the index. You can do this locally with pip index generate:

   # Generate index from your local temp directory
   pip index generate --directory ./temp_packages --output-file ./temp_packages/index.html
   

4. Upload the index.html file to the simple directory in your GCS bucket.
5. Install packages using the --index-url flag to point to your GCS index:

   pip3 install pandas --index-url https://storage.googleapis.com/public_bucket/simple/
   

Notes for Public Bucket Access #

• Ensure your GCS bucket and its objects are set to publicly readable (check the bucket permissions in the Google Cloud Console).
• If your bucket isn't fully public but you have authenticated access, run gcloud auth application-default login to let pip use your Google Cloud credentials to access the bucket.

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