浏览器的快速IPC方法

Question

I wrote an application that accumulates data (a few MB/sec, but updated like 10 times a second) and displays the current process in the browser via javascript.

The problem is that currently I write the data to file and load it with javascript, but this makes the application very laggy and people are complaining that their hdd is working a lot.

I would love to use some flags like "FILE_ATTRIBUTE_TEMPORARY" to tell my OS to not actually write the files to disk, but javascript requires me to close the file handle first (otherwise firefox couldn't open it). Thus it will be written to disk at that point, killing the point in initially using that flag.

I thought about using something like a mysql database, but I really want to keep it as simple as possible, and I would prefer a solution that doesn't force the user to set up some http or mysql server.

Answer 1

Include a webserver in your standalone C++ application and serve the data directly from memory.

I think that is the only option to avoid the I/O overhead which you are facing now. There is no interface to IPC, message queues or something similar in JS. TCP (or UDP, with websockets) seems to be the only possible way to avoid disk I/O. Another option would be to replace your JavaScript with a Browser Plugin which can access the capabilities of your operating system in native code.

Note: In a UNIX environment, you can create a FIFO socket in the File System, which may (or may not) serve your purpose as well. I don't think windows supports something like this, though.

Answer 2

It sounds like your application that accumulates data is running locally on the user's machine and that the page loaded into the browser is also local. As you stated you didn't want the user to have to configure (or I would assume run) an HTTP server, that you want to avoid writing the data to disk for perf reasons, and your title includes "IPC", I believe what you are left is creating a browser plug-in or including an ActiveX control in your page.

An ActiveX control in your page could utilize any of the IPC mechanisms available on the platform to native code (eg shared memory).

Another option would be to have your application be an HTTP server so a separate one was not needed, but that may be fall into the category of the user having to configure an HTTP server (they may need to open ports on a software firewall for example).

Answer 3

I would look into how to make a plug-in or a CGI program for your chosen webserver so that your Javascript can access a certain URL on your server and the data returned not come from a real file but just be served by your program handling the web request. (This is possible to do on most web servers. It could be as simple as your C++ program writing to standard console out.)