Download a private binary release file from GitHub in bash

download.sh

#!/usr/bin/env bash
#
# This is an adaptation of code I wrote to download a private binary from GitHub. Such...pain.
# Why can't GitHub just offer a standardized URL you can download a release binary from and attach 
# your Github Personal Access Token as a header? 
#
# Since this code is an adaptation it hasn't been directly tested, but the code it was adapted from works
# and hopefully you can get the missing piece you're after by looking here.
#

set -e

# Parse CLI args
readonly github_oauth_token="$1"
readonly git_tag="$2"
readonly github_repo_owner="$3"
readonly github_repo_name="$4"
readonly release_asset_filename="$5"
readonly output_path="$6"

# Get the "github tag id" of this release
github_tag_id=$(curl --silent --show-error \
                     --header "Authorization: token $github_oauth_token" \
                     --request GET \
                     "https://api.github.com/repos/$github_repo_owner/$github_repo_name/releases" \
                     | jq --raw-output ".[] | select(.tag_name==\"$git_tag\").id")

# Get the download URL of our desired asset
download_url=$(curl --silent --show-error \
                    --header "Authorization: token $github_oauth_token" \
                    --header "Accept: application/vnd.github.v3.raw" \
                    --location \
                    --request GET \
                    "https://api.github.com/repos/$github_repo_owner/$github_repo_name/releases/$github_tag_id" \
                    | jq --raw-output ".assets[] | select(.name==\"$release_asset_filename\").url")

# Get GitHub's S3 redirect URL
# Why not just curl's built-in "--location" option to auto-redirect? Because curl then wants to include all the original
# headers we added for the GitHub request, which makes AWS complain that we're trying strange things to authenticate.
redirect_url=$(curl --silent --show-error \
          --header "Authorization: token $github_oauth_token" \
          --header "Accept: application/octet-stream" \
          --request GET \
          --write-out "%{redirect_url}" \
          "$download_url")

# Finally download the actual binary
sudo curl --silent --show-error \
          --header "Accept: application/octet-stream" \
          --output "$output_path" \
          --request GET \
          "$redirect_url"

Probably don't need sudo on line 48. :)

I was starting pulling hairs from my head getting binaries from GitHub. Just removed the "sudo" from line 48. Thanks!

The first curl isn't necessary if you change the second curl to get the download url by tag
Change:
"https://api.github.com/repos/$github_repo_owner/$github_repo_name/releases/$github_tag_id" \
to:
"https://api.github.com/repos/$github_repo_owner/$github_repo_name/releases/tags/$git_tag" \

also, the first curl is paged by 30 releases per page. so if the $git_tag isn't found in the first page, that curl will return a jq error like this jq: error (at <stdin>:4): Cannot index string with string "tag_name" jq: error (at <stdin>:4): Cannot iterate over null (null) curl: (3) <url> malformed curl: (3) <url> malformed

ok. sorry. I have found another optimization. If you pass the filename via xargs, you can bypass that redirect problem. Other than defining the variable names, it becomes a one-liner and it'd look something like this:

curl -sL -H "Authorization: token $github_oauth_token" "https://api.github.com/repos/$github_repo_owner/$github_repo_name/releases/tags/$git_tag" | jq -r '.assets[] | select(.name == "'$release_asset_filename'").url' | xargs -I {} curl -sL -H "Authorization: token $github_oauth_token" -H 'Accept:application/octet-stream' -o $output_path {}

I made a new gist using this method here:
https://gist.github.com/mathew-fleisch/f2e0308eefe974701370f569027a7cfe

If anyone interest, give a try: https://github.com/zero88/gh-release-downloader

Actually, I can't believe I forgot to share here, but we wrote a first-class CLI tool in Go at https://github.com/gruntwork-io/fetch to do exactly what this script does and much more. It's worked well for us so far and seems to have a small, but active community of users.