Traversing with Pagination

The GitHub API provides a vast wealth of information for developers to consume. Most of the time, you might even find that you're asking for too much information, and in order to keep our servers happy, the API will automatically paginate the requested items.

In this guide, we'll make some calls to the GitHub Search API, and iterate over the results using pagination. You can find the complete source code for this project in the platform-samples repository.

Basics of Pagination

To start with, it's important to know a few facts about receiving paginated items:

  1. Different API calls respond with different defaults. For example, a call to list GitHub's public repositories provides paginated items in sets of 30, whereas a call to the GitHub Search API provides items in sets of 100
  2. You can specify how many items to receive (up to a maximum of 100); but,
  3. For technical reasons, not every endpoint behaves the same. For example, events won't let you set a maximum for items to receive. Be sure to read the documentation on how to handle paginated results for specific endpoints.

Information about pagination is provided in the Link header of an API call. For example, let's make a curl request to the search API, to find out how many times Mozilla projects use the phrase addClass:

curl -I "https://api.github.com/search/code?q=addClass+user:mozilla"

The -I parameter indicates that we only care about the headers, not the actual content. In examining the result, you'll notice some information in the Link header that looks like this:

Link: <https://api.github.com/search/code?q=addClass+user%3Amozilla&page=2>; rel="next",
  <https://api.github.com/search/code?q=addClass+user%3Amozilla&page=34>; rel="last"

Let's break that down. rel="next" says that the next page is page=2. This makes sense, since by default, all paginated queries start at page 1. rel="last" provides some more information, stating that the last page of results is on page 34. Thus, we have 33 more pages of information about addClass that we can consume. Nice!

Always rely on these link relations provided to you. Don't try to guess or construct your own URL.

Navigating through the pages

Now that you know how many pages there are to receive, you can start navigating through the pages to consume the results. You do this by passing in a page parameter. By default, page always starts at 1. Let's jump ahead to page 14 and see what happens:

curl -I "https://api.github.com/search/code?q=addClass+user:mozilla&page=14"

Here's the link header once more:

Link: <https://api.github.com/search/code?q=addClass+user%3Amozilla&page=15>; rel="next",
  <https://api.github.com/search/code?q=addClass+user%3Amozilla&page=34>; rel="last",
  <https://api.github.com/search/code?q=addClass+user%3Amozilla&page=1>; rel="first",
  <https://api.github.com/search/code?q=addClass+user%3Amozilla&page=13>; rel="prev"

As expected, rel="next" is at 15, and rel="last" is still 34. But now we've got some more information: rel="first" indicates the URL for the first page, and more importantly, rel="prev" lets you know the page number of the previous page. Using this information, you could construct some UI that lets users jump between the first, previous, next, or last list of results in an API call.

Changing the number of items received

By passing the per_page parameter, you can specify how many items you want each page to return, up to 100 items. Let's try asking for 50 items about addClass:

curl -I "https://api.github.com/search/code?q=addClass+user:mozilla&per_page=50"

Notice what it does to the header response:

Link: <https://api.github.com/search/code?q=addClass+user%3Amozilla&per_page=50&page=2>; rel="next",
  <https://api.github.com/search/code?q=addClass+user%3Amozilla&per_page=50&page=20>; rel="last"

As you might have guessed, the rel="last" information says that the last page is now 20. This is because we are asking for more information per page about our results.

Consuming the information

You don't want to be making low-level curl calls just to be able to work with pagination, so let's write a little Ruby script that does everything we've just described above.

As always, first we'll require GitHub's Octokit.rb Ruby library, and pass in our personal access token:

require 'octokit'

# !!! DO NOT EVER USE HARD-CODED VALUES IN A REAL APP !!!
# Instead, set and test environment variables, like below
client = Octokit::Client.new :access_token => ENV['MY_PERSONAL_TOKEN']

Next, we'll execute the search, using Octokit's search_code method. Unlike using curl, we can also immediately retrieve the number of results, so let's do that:

results = client.search_code('addClass user:mozilla')
total_count = results.total_count

Now, let's grab the number of the last page, similar to page=34>; rel="last" information in the link header. Octokit.rb support pagination information through an implementation called "Hypermedia link relations." We won't go into detail about what that is, but, suffice to say, each element in the results variable has a hash called rels, which can contain information about :next, :last, :first, and :prev, depending on which result you're on. These relations also contain information about the resulting URL, by calling rels[:last].href.

Knowing this, let's grab the page number of the last result, and present all this information to the user:

last_response = client.last_response
number_of_pages = last_response.rels[:last].href.match(/page=(\d+).*$/)[1]

puts "There are #{total_count} results, on #{number_of_pages} pages!"

Finally, let's iterate through the results. You could do this with a loop for i in 1..number_of_pages.to_i, but instead, let's follow the rels[:next] headers to retrieve information from each page. For the sake of simplicity, let's just grab the file path of the first result from each page. To do this, we'll need a loop; and at the end of every loop, we'll retrieve the data set for the next page by following the rels[:next] information. The loop will finish when there is no rels[:next] information to consume (in other words, we are at rels[:last]). It might look something like this:

puts last_response.data.items.first.path
until last_response.rels[:next].nil?
  last_response = last_response.rels[:next].get
  puts last_response.data.items.first.path
end

Changing the number of items per page is extremely simple with Octokit.rb. Simply pass a per_page options hash to the initial client construction. After that, your code should remain intact:

require 'octokit'

# !!! DO NOT EVER USE HARD-CODED VALUES IN A REAL APP !!!
# Instead, set and test environment variables, like below
client = Octokit::Client.new :access_token => ENV['MY_PERSONAL_TOKEN']

results = client.search_code('addClass user:mozilla', :per_page => 100)
total_count = results.total_count

last_response = client.last_response
number_of_pages = last_response.rels[:last].href.match(/page=(\d+).*$/)[1]

puts last_response.rels[:last].href
puts "There are #{total_count} results, on #{number_of_pages} pages!"

puts "And here's the first path for every set"

puts last_response.data.items.first.path
until last_response.rels[:next].nil?
  last_response = last_response.rels[:next].get
  puts last_response.data.items.first.path
end

Constructing Pagination Links

Normally, with pagination, your goal isn't to concatenate all of the possible results, but rather, to produce a set of navigation, like this:

Sample of pagination links

Let's sketch out a micro-version of what that might entail.

From the code above, we already know we can get the number_of_pages in the paginated results from the first call:

require 'octokit'

# !!! DO NOT EVER USE HARD-CODED VALUES IN A REAL APP !!!
# Instead, set and test environment variables, like below
client = Octokit::Client.new :access_token => ENV['MY_PERSONAL_TOKEN']

results = client.search_code('addClass user:mozilla')
total_count = results.total_count

last_response = client.last_response
number_of_pages = last_response.rels[:last].href.match(/page=(\d+).*$/)[1]

puts last_response.rels[:last].href
puts "There are #{total_count} results, on #{number_of_pages} pages!"

From there, we can construct a beautiful ASCII representation of the number boxes:

numbers = ""
for i in 1..number_of_pages.to_i
  numbers << "[#{i}] "
end
puts numbers

Let's simulate a user clicking on one of these boxes, by constructing a random number:

random_page = Random.new
random_page = random_page.rand(1..number_of_pages.to_i)

puts "A User appeared, and clicked number #{random_page}!"

Now that we have a page number, we can use Octokit to explicitly retrieve that individual page, by passing the :page option:

clicked_results = client.search_code('addClass user:mozilla', :page => random_page)

If we wanted to get fancy, we could also grab the previous and next pages, in order to generate links for back (<<) and forward (>>) elements:

prev_page_href = client.last_response.rels[:prev] ? client.last_response.rels[:prev].href : "(none)"
next_page_href = client.last_response.rels[:next] ? client.last_response.rels[:next].href : "(none)"

puts "The prev page link is #{prev_page_href}"
puts "The next page link is #{next_page_href}"