METAR cacheing / performance / code methods

[Phil P. <phi...@sw...>]

While I was trolling through the code, I often saw what looked like many 
and various chunks of code which essentially re-processed the RAW METAR 
data from scratch. (eg interpreting the timestamp).

[
Maybe I'm not being clear here... it looks to me like different bits of 
PHPWeather that need (for example) the timestamp re-interpret it from 
the RAW data themselves, rather than a chunk of code munging that into 
an internal datastructure (ie once) which all the code chunks read as 
they need.

Let me also just say that I've only done a little trolling around 
PHPWeather sources, and I'm barely able to 'crawl' when it comes to 
writing PHP. I'm not trying to criticize the code, merely wanting to 
better understand how it works (perhaps I might try my hand at 
contributing some tweaks/features in the not-too-distant future)
]

As this makes much and heavy use of regexes, I was wondering if this 
could have a material impact on performance.

So, I did some quick hacking around with my PHP scripts and came up with 
the following stats:

- full site page, including process (ensure this has been cached) METAR 
for TEXT as well as evaluate for ICONS)
	- 0.089 seconds or thereabouts
- same page, remove the code chunk/module for the METAR processing and 
output
	- 0.018 seconds

ugh! even when it's cached, processing the METAR and generating the 
html-ified output makes the page generation take 3 times longer than 
previously.

(of course, I'm using MySQL as the datastore for METARs, so this delay 
could be mostly due to MySQL query queueing and processing delays)

So I hacked around, and wrote some additions to my scripts to cache my 
formatted HTML output into a file. Now (as long as the cached results 
are within the timeout) showing the processed weather data takes negible 
additional time over normal page generation.

That's nice as a workaround. But obviously it only helps me and nobody 
else (well, not much).  Here's the code-fragment I used to 'cache the 
output HTML' if anyone else wants to use it.


/* LockFile must exist, cacheFile must be writeable by the webserver */

	if((time() - filectime($cacheFile)) < $metarCacheTimeout) {
		readfile($cacheFile);	
	} else {
		/* here we regenerate the formatted METAR results */

		$output = {munged html-ified output you want on your site};
	
		/* grab a lock on LOCKFILE, rewrite cachefile, release lock */
		$fp = fopen($cacheFile . '-LOCK', 'r');
		if( flock($fp, LOCK_EX + LOCK_NB) ) {
			$fp2 = fopen($cacheFile, 'w');
			if($fp2) {
				fwrite($fp2, $output);
				fflush($fp2);
				fclose($fp2);
			}
			flock ($fp, LOCK_UN);
		}
		echo $output;
	}

At some point I'll have a closer look and see if I can find where the 
bulk of the time is going, but for now I just thought I'd throw my 
findings out there and listen to any comments/feedback others might have.

Thanks,
Phil P