How Magic resolves URLs and endpoints

magic.endpoint is a dynamic endpoint URL controller, allowing you to declare endpoints that are dynamically resolved using your IHttpExecutorAsync service implementation. The default implementation of this interface, is the class called HttpExecutorAsync, and the rest of this file will be focused on documenting this implementation, since it is the default service implementation for magic.endpoint - Although, technically, you could exchange this with your own implementation if you wish, completely changing the behaviour of the library if you wish to for instance resolve endpoints to Python, Ruby, or any other dynamic programming language implementation, and you have some means to execute such code from within a .Net 6 environment. The resolver will be invoked for all relative URLs starting with “magic/”, for the following verbs.

The default service implementation will resolve everything after the “magic/” parts in the given URL, to a Hyperlambda file assumed to be found relatively beneath your “/files/” folder - Although, exactly where you physically put your files on disc, can be configured through your “appsettings.json” file. The HTTP verb is assumed to be the last parts of your filename, before its extension, implying an HTTP request such as the following.

GET magic/modules/foo/bar

Will resolve to the following physical file on disc.

files/modules/foo/bar.get.hl

Only the “magic” part of your URL is rewritten before the verb is appended to the URL, and finally the extension “.hl” appended. Then the file is loaded and parsed as Hyperlambda, and whatever arguments you pass in, either as query parameters or as your JSON payload is appended into your resulting lambda node’s [.arguments] node as arguments to your Hyperlambda file invocation. The resolver will never return files directly, but is only able to execute Hyperlambda files, so by default there is no way to get static files, unless you create a Hyperlambda endpoint that returns a static file somehow.

The default resolver will only allow the client to resolve files inside your “/files/modules/” folder and “/files/system/” folder. This allows you to safely keep files that parts of your system relies upon inside your dynamic “/files/” folder, without accidentally creating endpoints clients can resolve, resulting in breaches in your security. Only characters a-z, 0-9 and ‘-‘, ‘_’ and ‘/’ are legal characters for the resolvers, and only lowercase characters to avoid file system incompatibilities between Linux and Windows. There is one exception to this rule though, which is that the resolver will resolve files and folder starting out with a period (.) character, since this is necessary to allow for having “hidden files” being resolved as endpoints - Which is a requirement to make things such as Apple’s “.well-known” endpoints being resolved. Below is probably the simplest HTTP endpoint you could create. Save the following Hyperlambda in a file at the path of /modules/tutorials/foo.get.hl using for instance your Magic “Hyper IDE” menu item.

return
   result:Hello from Magic Backend

Then invoke the endpoint using the GET verb with the following URL.

http://localhost:5000/magic/modules/tutorials/foo

Arguments

The default IHttpExecutorAsync implementation can explicitly declare what arguments the file can legally accept, and if an argument is given during invocation that the file doesn’t allow for, an exception will be thrown and the file will never be executed. This allows you to declare what arguments your Hyperlambda file can accept, and avoid having anything but arguments explicitly declared in your Hyperlambda file from being sent into your endpoint during invocation of your HTTP endpoint. An example Hyperlambda file taking two arguments can be found below.

.arguments
   arg1:string
   arg2:int

strings.concat
   get-value:x:@.arguments/*/arg1
   .:" - "
   get-value:x:@.arguments/*/arg2

unwrap:x:+/*
return
   result:x:@strings.concat

If you save this file on disc as /files/modules/tutorials/foo2.get.hl, you can invoke it as follows using the HTTP GET verb - Assuming your backend is running on localhost at port 5000.

http://localhost:5000/magic/modules/tutorials/foo2?arg1=howdy&arg2=5

JSON payloads and form URL encoded payloads are automatically converted to lambda/nodes - And query parameters are treated indiscriminately the same way as JSON payloads - Except of course, query parameters cannot pass in complex graph objects, but only simply key/value arguments. Only POST, PUT and PATCH endpoints can handle payloads. If you supply a payload to a GET or DELETE endpoint, an exception will be thrown, and an error returned to the caller.

To allow for any arguments to your files, simply ommit the [.arguments] node in your Hyperlambda althogether, or supply an [.arguments] node and set its value to *. Alternatively, you can also partially ignore arguments sanity checking of individual nodes, by setting their values to *, such as the following illustrates.

.arguments
   arg1:string
   arg2:date
   arg3:*

In the above arguments declaration, [arg1] and [arg2] will be sanity checked, and input converted to string or date (DateTime) - But the [arg3] parts will be completely ignored, allowing the caller to invoke it with anything as arg3 during invocation - Including complete graph JSON objects, assuming the above declaration is for a PUT, POST or PATCH Hyperlambda file. The ‘*’ value for an argument also turn off all conversion, implying everything will be given to your lambda object with the JSON type the argument was passed in as. All arguments declared are considered optional, and the file will still resolve if the argument is not given, except of course the argument won’t exist in the [.arguments] node. However, no argument not found in your [.arguments] declaration can be provided during invocations, assuming you choose to declare an [.arguments] collection in your Hyperlambda endpoint file, and you don’t set its value to *.

To declare what type your arguments can be, set the value of the argument declaration node to the Hyperlambda type value inside of your arguments declaration, such as illustrated above. Arguments will be converted if possible, to the type declaration in your argument’s declaration. If no conversion is possible, an exception will be thrown. Although the sanity check will check graph objects, passed in as JSON payloads, it has its restrictions, such as not being able to sanity check complex objects passed in as arrays, etc. If you need stronger sanity checking of your arguments, you will have to manually check your more complex graph objects yourself in your own Hyperlambda files.

Also realise that if the value originates from a payload, as in from a PUT, PATCH or POST JSON object for instance, these types of objects might contain null values. If they do, no conversion will be attempted, and internally within your endpoint’s Hyperlambda code, you might therefor expect to see for instance long values being in fact null, even though technically these are not nullable types in .Net.

Accepted Content-Type values

The POST, PUT and PATCH endpoints can intelligently handle any of the following Content-Types.

JSON types of payloads are fairly well described above, and URL encoded form payloads are handled the exact same way, except of course the [.arguments] node is built from URL encoded values instead of JSON - However, internally this is transparent for you, and JSON, query parameters, URL encoded forms, and “multipart/form-data” can be interchanged 100% transparently from your code’s perspective - Except “multipart/form-data” might have [file] arguments wrapping streams that you need to handle separately as such. File attachments will be passed into your endpoint as follows.

.arguments
   file
      name:filename-on-client.txt
      stream:[raw Stream object here]

All other types of payloads will be passed in as the raw stream, not trying to read from it in any ways, allowing you to intercept reading with things such as authentication, authorisation, logic of where to persist content, etc. To understand how you can handle these streams, check out the “magic.lambda.io” project’s documentation, and specifically the [io.stream.xxx] slots.

Extending the Content-Type request and response resolver

The Content-Type resolver/parser is extendible, allowing you to change its behaviour by providing your own callback that will be invoked for some specific Content-Type value provided. This is useful if you want to be able to for instance handle “text/xml” or “text/csv” types of request/response objects, and intelligently and automatically create an argument collection from it. Below is example code illustrating how to create your own HTTP request resolver for the MIME type of “application/x-foo”.

EndpointController.RegisterContentType("application/x-foo", async (signaler, request) =>
{
   var args = new Node();

   /* ... Create some sort of collection of arguments and put into args node here ... */

   return args;
});

Notice - The argument sanity checking will still be invoked with a custom handler, implying your Content-Type handler and the [.arguments] declaration in your Hyperlambda file still needs to agree upon the arguments, and if a non-valid argument is specified to a Hyperlambda file, an exception will be thrown. Also notice that registering a custom Content-Type is not thread safe, and should be done as you start your application, and not during its life time. You can also provide your own HTTP response resolver that will be invoked given some specified Content-Type from your Hyperlambda file. This is done in a similar manner using something resembling the following.

EndpointController.RegisterContentType("application/x-foo", (response) =>
{
   /* ... Return some sort of IActionResult here ... */
   return new ObjectResult(response.Content) { StatusCode = response.Result };
});

The above method should also exclusively be used during startup, and not later, since it is not thread safe. The above method assumes you register your Content-Type handlers as your application is starting.

Meta information

Due to the semantic structure of Hyperlambda, retrieving meta information from your HTTP endpoints using this module is very easy. The project has one slot called [endpoints.list] that returns meta information about all your endpoints. This slot again can be invoked using the following URL.

http://localhost:5000/magic/system/endpoints/list

This endpoint/slot will semantically traverse your endpoints, recursively loading up all Hyperlambda files from disc that are resolved from a valid URL, and return meta information about the file/endpoint back to the caller. This allows the system to easily figure out things such as the following about your endpoints.

This slot/endpoint is what allows you to see meta information about all your HTTP REST endpoints in the “Endpoints” menu item in the Magic dashboard for instance. The return value from this slot/endpoint again, is what’s used as some sort of frontend is being generated using the Magic dashboard.

Extending the meta data retrieval process

You can extend the meta data retrieval process by invoking ListEndpoints.AddMetaDataResolver, and pass in your own function. This class can be found in the magic.endpoint.services.slots namespace. The AddMetaDataResolver method takes one function object, which will be invoked for every file the meta generator is trying to create meta data for, with the complete lambda, verb and args of your endpoint. This allows you to semantically traverse the lambda/args nodes, and append any amount of (additional) meta information you wish - Allowing you to extend the generating of meta data, if you have some sort of general custom Hyperlambda module, creating custom HTTP endpoints of some sort.

This function will be invoked for every single Hyperlambda file in your system, every time meta data is retrieved, so you might want to ensure it executes in a fairly short amount of time, not clogging the server or HTTP endpoint meta generating process in any ways.

In addition to the meta retrieval endpoint described above, the module contains the following slots.

Misc

Unless you explicitly change the Content-Type of your response object, by using the [response.headers.set] slot, a Content-Type of application/json will be assumed, and this header will be added to the resulting HTTP response object. If you wish to override this behavious and return plain text for instance, you could create an endpoint containing the following.

response.headers.set
   Content-Type:text/plain
return:Hello from Magic Backend

If you intend to return anything but JSON, you must set the Content-Type header, because the resolver will by default try to serialize your content as JSON, and obviously fail unless it is valid JSON. You can also return stream objects using for instance the [return-value] slot, at which point ASP.NET Core will automatically stream your content back over the response object, and Dispose your stream automatically for you afterwards. This allows you to for instance return large files back to the client without loading them into memory first. If you do this, you’ll have to change your Content-Type accordingly.

Cookies

Since cookies have more parameters than just a simple key/value declaration, the [response.cookies.set] slot takes the following arguments.

Only the [value] from above is mandatory. To delete a cookie on the client, set the expiration date to a value in the past.

Project website

The source code for this repository can be found at github.com/polterguy/magic.endpoint, and you can provide feedback, provide bug reports, etc at the same place.

Quality gates