Azure DevOps feed get download url for a latest version of the nuget package or symply download latest package - azure-devops

Question

I've got the Azure DevOps pipeline which builds nuget package and deploys it to Azure DevOps feed. In the portal I can find download link to a specific version of the package.
How do I find url to download latest version of the nuget package in the feed?
Alternatively how do I download the latest version of the package?
Ideally just via curl, dotnet or anything what is resent on dev windows machine and in general docker sdk image.
I tend to go long way
dotnet new console
add package
restore
find the location of the file. But I really don't like this approach. Anything nicer?

Answer 1

How do I find url to download latest version of the nuget package in the feed?
Please follow below steps to find this url.
Use this Rest API: Feed Management - Get Feeds to get the feed
id.
Use this API: Artifact Details - Get Packages to get details about the target package in the feed. Url looks like: https://feeds.dev.azure.com/{organization}/{project}/_apis/packaging/Feeds/{feedId}/packages?packageNameQuery={packageName}&api-version=6.0-preview.1. From the response, you will find its versions array, and the latest version is marked by "isLatest":true, so you get the latest version of this package in this feed.
This Rest API: NuGet - Download Package provides the url to download latest version of the nuget package in the feed.
BTW, please note that the project parameter in above APIs must be supplied if the feed was created in a project. If the feed is not associated with any project, omit the project parameter from the request.

Answer 2

If you build the latest before you import NuGet packages and set the project(s) up to get latest, you can make this automagic. Neither of these are really part of pipeline, except that you are ordering steps.
But, I question NuGetting a project and then including in pipeline, especially if this is one project (with dependencies) in a larger solution. If you are to the point you are deploying to NuGet, you should be more intentional and move the items in question to another solution and productize the solution rather than leave it in a master solution. You would then create a separate pipeline for the new product and consume it just like any other NuGet package. Now, you may think something like "but I use this in a variety of solutions" so this is easier, but, in reality, that is a more compelling reason to separate it out and have it intentionally placed in NuGet and not automatically get latest (i.e. act if you are buying this assembly from another company and have governance that requires you test the new version before automatically deploying with your solutions).
If you are doing this because the project(s) in question are still in flux, then I would not set the consumer to automatically pick up latest and make it intentional. Even if you currently only have a single development group, you are best to commoditize the parts going to NuGet. Otherwise, there is really no need to build a NuGet package and consume (unless I am missing some compelling reason not to productize and continue this complex manner of compiling each time and versioning).
Probably TL;DR?

Answer 3

For anyone who found #edward-han-msft's answer useful (as I did) here's a python 3.6+ script to download all versions of all packages from an Azure DevOps artifact(sic) feed. In my example, I was migrating to another NPM feed so this script also publishes the downloaded packages to whatever npm registry is configured in your .npmrc. Adjust it to your needs.
import requests
from requests.auth import HTTPBasicAuth
import re
import json
import os
from os.path import exists
import subprocess
organization = '<Your organisation>'
project = '<Your project name>'
feed_name = '<The name of the artifact feed>'
feed_id = '<feedId - this can be found by examining the json response of the package feed>'
# Packages feed url
url_packages = f"https://feeds.dev.azure.com/{organization}/{project}/_apis/packaging/feeds/{feed_name}/packages?api-version=5.1-preview.1"
# ADO PAT
basic = HTTPBasicAuth("<I think this can be anything>", "<a DevOps PAT with Packaging Read scope>")
# fetch the packages feed and save locally
r1 = requests.get(url_packages, auth = basic)
open('packages.json', 'wb').write(r1.content) # for debug
# parse the json
packages = json.loads(r1.content)
for package in packages['value']:
package_name = package['normalizedName']
short_name = package_name.split('/')[1]
versions_url = package['_links']['versions']['href']
print(f'Package: {package_name} ({short_name})')
# create a folder for each package
package_folder = './'+short_name
if not exists(package_folder):
os.mkdir(package_folder)
# fetch versions json
r2 = requests.get(versions_url, auth = basic)
versions = json.loads(r2.content)
open(f'{package_folder}/versions.json', 'wb').write(r2.content) # for debug
# This block iterates though the versions and discards ones that fall outside of semver e.g. x.x.x-canary or similar
version_numbers = {}
for package_version in versions['value']:
# is it a release version? (semver compliant)
version_number = package_version['normalizedVersion']
match = re.search('^\d+.\d+.\d+$', version_number)
if match:
split = version_number.split('.')
sortable_version = '%02d.%02d.%02d' % (int(split[0]),int(split[1]),int(split[2]))
version_numbers[sortable_version] = version_number
# the dictionary keys are a sortable format of the version e.g. 00.00.00
version_numbers = sorted(version_numbers.items(), key = lambda kv: kv[0])
print(version_numbers) # for debug
for package_version in version_numbers:
version_number = package_version[1] # package_version is a tuple
package_filename = f'{package_folder}/{short_name}-{version_number}.tgz'
# dowload package if not previously downloaded
if not exists(package_filename):
print(f'Downloading : {short_name}-{version_number}')
package_url = f'https://pkgs.dev.azure.com/{organization}/{project}/_apis/packaging/feeds/{feed_id}/npm/packages/{package_name}/versions/{version_number}/content?api-version=6.0-preview.1'
r3 = requests.get(package_url, allow_redirects=True, auth = basic)
open(package_filename, 'wb').write(r3.content)
# publish the package if not previsouly published
if not exists(package_filename+'.published'):
npm_publish = subprocess.run(["npm", "publish", package_filename])
if npm_publish.returncode == 0:
# create a file matching the package with .published extension to indicate successful publish
subprocess.run(["touch", package_filename+'.published'])
else:
print(f'Error publishing {short_name}-{version_number}. Code: {npm_publish.returncode}')
print('done.')
The script is idempotent as it keeps a copy of downloaded packages and also touches a <package name>.published file after a successful call to npm publish.