I am working with a third-party web service that was originally integrated into a .NET application. I am looking to port this functionality over to an elixir web application. The company that provides the web service does not provide a WSDL so I can’t use a SOAP library like elixir-soap/soap.
Below is an example SOAP request that is made. Let’s say the request gets the inventory for a store.
Question 1: Is phoenix the best tool for something like this?
Question 2: What the best way to build these XML SOAP requests in elixir prior to sending?
Question 3: What is the best way to send these requests via HTTP?
Question 4: What is the best way to parse the responses?
Q1: Phoenix is a web server, not an HTTP request client, but there is no reason not to be able to integrate that.
Q2: I can’t answer that because I don’t use SOAP, but a simple Google search for: “soap elixir” should give you a decent start. You can always build it yourself if you need it to be custom. In the end, it is just an HTTP request.
Q3: HTTP client libraries and wrappers
Q4: Are you parsing a JSON or XML?
Thanks arcyfelix! I get that phoenix is not an http client. Just looking for personal opinions on whether or not phoenix is the right choice for a task like this over languages like C# or Java.
I would approach the problem a little bit differently by asking two questions:
What are the subparts of my problem?
In your case: A web server, a soap library, an HTTP client and an xml parser.
What are the best tools for the job?
Well, I generally just Google: “<my problem / subsystem> elixir” or search in the forum and I look at if the project is maintained or complete.
Really hard to answer without seeming biased but, with respect to your requirements, I think Phoenix is indeed one of the best tools for your project, because:
You essentially need a reverse proxy of sorts, a rewriting one i.e. you accept certain data, massage it and send it downstream to other services. And then get data back from them, maybe massage them as well, and return those to your users.
The Erlang’s runtime is excellent at this because if your downstream services are slow – say, it needs 5-10 seconds to respond – then your Phoenix app will easily take it because you have hundreds of thousands of such requests waiting and every single one of them will get fair treatment. You only have to tune your own Phoenix app’s response timeouts (i.e. give them more generous values, from 10s and up I suppose).
Your memory load should remain predictable because each request (== Erlang “process” which is NOT the same as OS process!) is isolated and garbage-collected separately.
All in all, your latency will be predictable, memory requirements and pressure will be stable, and 99.9999% of all your problems will be with your downstream services and not with Phoenix (unless you get DDoS-ed).
I would say reach for Erlang’s :xmerl_scan directly but be warned: it’s not trivial. You can make some scary fast code, but it’s not easy to write. There are several libraries i.e. sweet_xml that step on top of xmerl that might be a bit easier for you but be warned (again) that you have to monitor your app well – good telemetry and metrics will be pretty much mandatory so you can keep track of CPU and memory usage and be able to switch to another library if the current one is under-performing or is straining your hosting too much.