The Application Programming Interface for this web application is designed according to the RESTful architectural style. Any client program or software that can speak HTTP can therefore be used to interact with the system.
In this tutorial we will use Python and its standard module httplib. This is a very basic HTTP interaction module. It has the advantage of showing explicitly the mechanism of HTTP interactions.
Obtaining a document from the web service
Let us obtain a document given a known URL. In the Python script tut1.py, we simply get the top page for the Adhoc web site:
This example illustrates several points:
A peculiar quirk of the httplib package is that once a connection instance has been used for a single request/response cycle, it becomes useless. One must create a new connection instance to perform another request.
Requesting a particular representation format
There are two ways of requesting a particular format of the representation:
The first method is more general and HTTP-ish. As in the Python script tut2.py:
We tell the server that we can accept only JSON format by specifying the mimetype 'application/json' in the 'Accept' header of the request, as defined by HTTP.
The representation is sent as a JSON data structure. We verify this by looking at the 'Content-Type' header of the response.
We convert the JSON representation to a Python data structure using the 'loads' function of the standard json module. We happen to know that it is a dictionary, and we therefore print the keys of that dictionary to have a look at what we got.
Adding login information to a request
The previous requests were done without login, so the Adhoc system viewed these as done under the account 'anonymous'. For most activities in Adhoc, you will want to login to your own account. Thus, authentication is needed for login.
Due to the stateless nature of the HTTP protocol, the authentication data (user account and password) has to be included in every request.
The Adhoc system currently uses only Basic Authentication. In cases where security is an important issue, TLS should therefore be used (i.e. the 'https' scheme, which is HTTP over an encrypted connection).
HTTP Basic Authentication works by packaging the user account name and the password together into a base64 message, and adding this as a header to the HTTP request, as in the Python script tut3.py:
Try this with the wrong password, and you will get the response status 401 'Unauthorized'.
Creating a task
Now for some real work. We want to use BLAST to search a protein query against a protein database. The blastp tool does that.
The basic idea with the Adhoc web service is that one does a HTTP POST to the chosen tool with all required input data in an appropriate representation. The Adhoc system creates a task and launches it, and responds by sending a HTTP Redirect containing the URL of the newly created task. The task is executed on the server asynchronously, and the client script will have to poll the task now and then until it has finished.
In order to figure out what input is required for a tool, look at the page for that tool using an ordinary browser. The input parameters for the tool are described there.
For the exact names of the input parameters, one can look at the appropriate section in the documentation page for the API.
One should also look in the TXT representation of the tool page, since the parameters and the allowed options are listed there.
For the blastp tool we need to specify the database(s), the query sequence in FASTA format (although the server is forgiving, and also accepts plain text without the proper FASTA header), the E-value cutoff and the output format. We encode the data into JSON and send this as the body of the request.
In the Python script tut4.py the task is created by the POST request. Its URL is defined by the UUID given to the task on creation.
We then loop to check what the status of the task is, polling every second. We really ought to check whether the task failed, but in this example we don't, for simplicity.
When the task is finished, we obtain the output by sending a GET request to the 'output' sub-resource of the task. We write this into the file tut4_output.txt.
Deleting a task
A task is persistent. It remains on the server until it is explicitly deleted. The Python script tut5.py shows how a task can be deleted. The URL of the task must be known in order to delete it.
The response for the DELETE request is a HTTP Redirect to the list of tasks for the account owning the deleted task.
|Adhoc 12.7||SciLifeLab tools||Per Kraulis (firstname.lastname@example.org)|