Share file caching across markbind instances

[nicholaschuayunzhi]

markbind/lib/parser.js

Lines 73 to 76 in eed11dd

	try {
	self._fileCache[filePath] = self._fileCache[filePath] ?
	self._fileCache[filePath] : fs.readFileSync(filePath, 'utf8');
	} catch (e) {

Currently each Parser instance has its own cache. Each Page creates its own parser or markbinder instance and thus the cache is not shared. Considering there is a lot file reuse in Markbind, I feel we should have a common cache to cut down on the IO operations.

I propose having a static instance of file cache for Parser class. Possibly can abstract as its own module?

Things to consider

Not reading from cache when file is changed during live reload.

Related #88

[nicholaschuayunzhi]

Update

Before (Build time, number of cache hits)

• build: 34344.112ms, 5143
• build: 36372.435ms, 5143
• build: 34837.442ms, 5143

After

• build: 33073.394ms, 13563
• build: 26338.988ms, 13563
• build: 29040.912ms, 13563

https://github.com/nicholaschuayunzhi/markbind/tree/138-static-file-cache
https://github.com/nicholaschuayunzhi/markbind-cli/tree/138-static-file-cache

Waiting for consensus for #139

[damithc]

@nicholaschuayunzhi can we resume the work on this issue?. Any performance gain in the build is a worthy cause.

[nicholaschuayunzhi]

Sorry for the late response, I've been trying this out for a bit.

While there is an increase in Cache Hits, there does not seem to be improvements in memory and performance. I'm not too sure of the reasons why - perhaps IO operations is not bottleneck of the build process? Tested on 2103 website-base

Before

heapUsed 151.83 MB
Cache Hit: 12510
Cache Miss: 10793
Build Time: 100524.524ms

After

heapUsed 156.57 MB
Cache Hit: 21823
Cache Miss: 1480
Build Time: 114969.759ms

I've made the changes here: nicholaschuayunzhi@83aa07d

[damithc]

While there is an increase in Cache Hits, there does not seem to be improvements in memory and performance.

Thanks for checking @nicholaschuayunzhi. I guess we can leave it be if there is no noticeable performance gain.