Did you know ... Search Documentation:
Processing HTTP-parameters

In HTMLRules, we used the HTML generation library to build a single page dynamically. In this page we will introduce two new primitives:

  1. Create a handler that processes parameters
  2. Create a URI to another handler dynamically

Creating links to another handler with parameters

First, the overall skeleton. This is very much like the HTMLRules example, but we introduce new rules for making a link to a module and including the file-name of a module. Introducing such tiny rules and managing a pool greatly boosts reusability in and maintainability of the code. Started using ?- server(5000)., the server is accessible at http://localhost:5000/

:- use_module(library(http/thread_httpd)).
:- use_module(library(http/http_dispatch)).
:- use_module(library(http/html_write)).
:- use_module(library(http/http_parameters)).	 % new
:- use_module(library(uri)).			 % new

:- http_handler(root(.),      list_modules, []). % /
:- http_handler(root(module), list_module,  []). % /module?name=<module>

server(Port) :-
        http_server(http_dispatch, [port(Port)]).

%%	list_modules(+Request)
%
%	Create a table of all available modules with their source-file

list_modules(_Request) :-
        findall(M, current_module(M), List),
        sort(List, Modules),
        reply_html_page(title('Loaded Prolog modules'),
                        [ h1('Loaded Prolog modules'),
                          table([ \header
                                | \modules(Modules)
                                ])
                        ]).

header -->
        html(tr([th('Module'), th('File')])).

modules([]) -->	[].
modules([H|T]) -->
        html(tr([td(\module_link(H)), td(\module_file(H))])),
        modules(T).

module_file(H) -->
        { module_property(H, file(Path)) }, !,
        html(Path).
module_file(_) -->
        html(-).

Now comes ones of the new parts. Instead of including the name of the module, we must create a link that, when clicked, invokes the list_module(+Request) handler. We can do this using the code below. This example introduces two new primitives:

  1. Attribute-values can be of the form A+B, which evaluates to string-concatenation.
  2. Attribute-values can be of the form encode(X), which evaluates to %-encoding of X
        ...,
        html(a(href('/list_module?name='+encode(Module)))

Although this type of coding is popular, it is not ideal because the location of /list_module is hard-coded. We do not care about the HTTP location, we want to call list_module(+Request)! The HTTP infrastructure can tell us this location and provides the predicate http_link_to_id/3 to compute a link based on the identifier of the handler (by default the predicate name) and additional parameters.

module_link(H) -->
        { http_link_to_id(list_module, [name=H], HREF) },
        html(a(href(HREF), H)).

Note that by using the URI encoding and decoding library(uri), we are sure that the generated HREF is properly encoded. The last step is to define our handler for /list_module?name=<module>.

Processing parameters

Dealing with parameters is achieved through http_parameters/2. The first argument is the Request passed by the server-library. The second argument is a list of Name(ValueVar, Properties). The properties allow for specifying the type, whether the parameter is optional, a default, etc. See http_parameters/2.

%%	list_module(+Request)
%
%	List info for a given module.

list_module(Request) :-
        http_parameters(Request,
                        [ name(Module, [])
                        ]),
        module_public_predicates(Module, Preds),
        reply_html_page(title('Module ~w'-[Module]),
                        [ h2('Public predicates for module ~q'-[Module]),
                          \predicates_ul(Preds)
                        ]).

%%	predicates_ul(+Module)// is det.
%
%	Generate an HTML =ul= list of public predicates in Module.

predicates_ul(Preds) -->
        html(ul(\predicate_list(Preds))).

predicate_list([]) --> [].
predicate_list([H|T]) -->
        html(li('~q'-[H])),
        predicate_list(T).


%%	module_public_predicates(+Module, -PublicPreds)

module_public_predicates(Module, Preds) :-
        findall(H, predicate_property(Module:H, exported), HL),
        maplist(head_to_pi, HL, Preds0),
        sort(Preds0, Preds).

head_to_pi(Head, Name/Arity) :-
        functor(Head, Name, Arity).

The remainder of the code is similar to listing the modules. However, we do introduce some reusability principles:

  1. Create a predicate for each interesting computation (in this case generate a list of public predicates).
  2. Create a rule for reusable UI components. predicates_ul//1 writes an UL-list of predicates, something which is likely to be reusable in this application (list undefined, private, etc. predicates).

Summary

In this page we learned linking dynamically generated pages together and passing parameters. The same infrastructure can of course be used to generate forms as we discuss in HTMLForms. We also learned how to deal with reusability and maintenance by properly splitting the code and by referring to HTTP paths by the predicate that implements them rather than the exact location of the server.

What is still missing?

If you loaded the source and ran the demo, you might feel a bit disappointed: it works flawlessly, but it looks very 1980's: totally basic style and no fancy interaction. With the technology presented so far, we can easily make it look 1990's: create reusable rules that create tables-in-tables(-in-frames)! In HTMLStyle, we discuss how to do this state-of-the-art.

See also
- The source: modules.pl
- Next: HTMLStyle.txt