As the subject says, I'm looking for a way to notify the caller of a time consuming process.

An example could be this:

procedure TMyHandler.TimeConsumingProcess(Sender: TObject; ARequest:
TRequest; AResponse: TResponse; var Handled: Boolean);
var
  I: Integer;
begin
  for I := 0 to 1000 do
  begin
    AResponse.Content := IntToStr(I);
    Sleep(1000);
  end;
  Handled := True;
end;

Of course this doesn't work. How can I send responses by intervals?.

Regards,

How can I send responses by intervals?

AFAIK that's not the way web application works. It's the client that should be querying the server in a regular interval (using AJAX request perhaps). The server can track current progress status and return that right away when asked.

Yes, I know that, but I thought Lazarus have something easy to develop such process.

The way I'm thinking of implementing is this:

1) The client app asks the server (a CGI process) for a process ID, for example 123, by calling getProcId.
2) Using this ID, the client calls the time-consuming task, passing the ID. for example, runTask(Id). At this time, the server launches a program passing the ID for the task. This must be a multi-thread program, because it can be asked for running many tasks, and can be asked for the status of each task. The program only finishes when the last task is ended.
3) The client calss a getProcStatus(ID) method. The server, using IPC or Pipes, asks the already-launched process for its status.

Regards,

Your program is the main thread. For each request your program should create a new thread to process. Each thread synchronizes with the main thread (your program) about it your own status.

Already exists an example with the same idea (more or less)... in Pascal... called ExtPascal.

Marcos Douglas

I have code which does exactly that. It's used in production. It's unix-only, however (it relies on fork and exec).

If you're interested, I can send you the code.

Michael.

AFAIK, a web application uses the plain old standard mechanism, a web server uses to work with a CGI application. it start the application and when same ends, the web server retrieves its output and sends it to the browser. So the web application just does not live long to be able to wait for anything.

To create a longer living web enabled process it takes a lot more effort.

You can try several approaches:

- purely propriety: Use a second process (a long living daemon) and have the web application communicate with same (e.g. via TCP/IP or Pipe)
- fast CGI: here the (Apache) Web Server does exactly this communication on it's own account
- ISAPI: here a (Microsoft) Web server communicates with a DLL

My colleagues do a combination of (1) and (3) with one of their (Delphi) project very successfully:

They created a very simple ISAPI DLL (using RemObjects to do the ISAPI communication)

They created a Windows service and used RemObjects (using the "Windows Message" transport) to have the ISAPI DLL communicate with the service.

(RemObjects is a commercial product that is specified to work as well with Delphi as with FreePascal.)

-Michael

I have code which does exactly that. It's used in production. It's unix-only, however (it relies on fork and exec).

GREAT !

It would be great if this would be integrated in the Lazarus distribution (as a special "Application") or if it would be available as an add-on package.

Thanks,
-Michael

Thanks Michael, I'm testing a different approach, based on a temporary file. It works this way:

1) The client asks for a TaskId, for example 1234.
2) With the TaskId generated by the server, the client calls a CGI method called runLongTask(myTaskId). This method writes a temporary file which name is the taskId.
3) The client can ask for status by using a method called getTaskStatus(myTaskId).

This way I can run a long task without having to create a daemon.

Regards,

Yes please, send me the code.

Regards,

That is exactly what I do.

Michael.

Sent in private mail.

Michael.

Leonardo M. Ram? wrote

2) With the TaskId generated by the server, the client calls a CGI method called runLongTask(myTaskId).

I don't know if it's a goad idea to allow a standard CGI to do a looooong action before returning top the WebServer.

At least the WebServer might kill the child task after a few seconds if it thinks it has run wild.

Maybe it is possible to configure the appropriate Timeout in the web server parameters.

-Michael

+1

IMHO, maybe is better to use CGI gateway to capture the requests and, for each one them, start a process -- this method will use a file to write the log of process, as Leonardo and Michael said.

Marcos Douglas

Marcos Douglas wrote

IMHO, maybe is better to use CGI gateway to capture the requests and, for each one them, start a process -- this method will use a file to write the log of process, as Leonardo and Michael said.

Yes, in my case, I know in advance the process wouldn't take more than one minute, but for longer processes the gateway is a nice idea.

Regards,

Well, I use the code I sent to take backups or fix databases, which takes 30 minutes easily. The browser just polls ever so-and-so time.

Michael.

Leonardo M. Ram? wrote

in my case, I know in advance the process wouldn't take more than one minute,

I feel that (without some special configuration) a normal web server will kill a standard CGI process that takes more than just a few seconds before returning.

Please check the specs of your Webserver and let us know what you find.

-Michael

Apache defaults to 300 seconds for its core timeout (which includes mod_cgi). Different modules have their own configuration settings, e.g. php have a max_execution_time, which defaults to 30 seconds afaik.

Henry