php code size limit?

ksandom

I'm working on code that generates code. The generated code works, except when I build on larger amounts of data it gives up. When the generated file is at about 800-900KB it works fine. When the file is about 1.5MB, it gives up.

When it gives up:
The browser downloads the original generated file completley uninterpreted. The problem is that:
1) The browser shouldn't download the file to disk, it should display it. (It does work fine for smaller amounts of data).
2) The server should be interpreting it before spitting it out. (Once again, it works fine on smaller files.)
3) It doesn't give any form of error (none displayed, and none in the file downloaded)

I am aware that php has a memory size limit per script being run, and I can lookup how to modify this if need be (I don't think this is worthwhile modifying the server environment since this util is likley to run on many setups). I don't think I am reaching this yet though. I can make an effort to make the file size smaller, however the gain would be minimal and there could be larger sizes in data later on anyway.

What I'm wanting to know is:
Does anyone know any causes of a script to dump instead of run? (sounds a bit sick doen't it? 😉 )

Any thoughts would a very welcome. Thanks 🙂

drawmack

You're probably running into server process time outs here. The server is what's giving up, not PHP.

Let me explain this too you.

For this discussion, there are two types of programming languages interpreted and compiled (we won't get into those that attempt to fall somewhere in the middle like Java).

In a compiled language you write the code, run the code through a compiler which then creates a file that is in native machine language.

In an interpreted language you write the code, then when you wish to run the program you run it through the intrepeter which turns it into native machine code on the fly and executes it without saving the output.

This means that all code is basically compiled. Interpreted code is compiled when it's run, every time it's run. Compiled code is compiled once and then run over and over.

The upside of interpreted code is that code changes are easier to implement cause there are no intermediate steps for the programmer.

The downside is that it takes A LOT longer for the code to begin executing.

Now how does this relate to your problem you ask?

Well, if the file is 1.5 megabytes and each character is 1 byte you're looking at about 1572864 characters in that file. So the php interpreter has to parse over 1 million characters turn it into machine code and then execute it. The server is probably nailing this as a run away process and just sending out the raw code. Now since the raw code is sent with a header of text/php and the browser doesn't know what to do with text/php it tries to save the file to disk.

ksandom

If I recall correctly, the default timeouts are between about 60-120 seconds depending which timeout you are talking about (eg httpd, php, browser).

The entire action happens in about 5 seconds including download (it's all local at the moment (but has also been tested externally)). I've done quite a bit with very time consuming tasks under php using such functions as set_time_limit.

I thankyou for the suggestion, but I think it is unlikley here. The response is sufficiently quick, that it appears to be making a descision. I trying to find out what that descision is. Originally I was thinking about memory usage and file size. But I'm also now wondering about logical complexity.

Does anyone know how many levels of nesting can occur? I'm estimating that the program is currently using about 9-10 levels of nesting with the potential to have more. Currently the nesting is in IFs. I recall that windows used to have a stack size of over 9000 by default. I suspect it's bigger now. I'm on linux and I'm not sure what the stack size of the operating system is or if it is even static. Having said that, I would have thought that 9-10 levels would not be much of what ever is remaining.

drawmack

Look in your server config files for something that sets a size limit on php files.

ksandom

Looked in config files/php info etc. Didn't find anything I didn't already know.
I've checked the memory usage, the working scripts are about 1-3MB.

I made a small script to call the non-working one as an include (but outputting a small amount of text first). The script runs, but the include behaves the same (it doesn't go).

Some more information. When the non-working script is run, no http headders are sent, but they are when it is run as an include. - No surprise, but it may give a clue as to where the descision is being made.

I've tried multiple browsers to rule out the possibility of a browser quirk.

Another possibility, the test machine is running an older version of php/apache/linux than the production machines. It's possible that the problem is comming a limitation/bug/feature that has changed in recent versions.

At the moment, I have enough functionality to keep moving and I'll come back to the larger file sizes when I have more time. At which point I'll do some more testing to see if I can narrow down the problem.

When I have had a chance to experiment more, I'll post again with any theories or results. In the mean time, if anyone has any more suggestions, they would be very welcome.

Thanks
- Kevin 🙂