Bad performance: Many references

[mity]

Processing many reference definitions/references takes too much time. On my machine:

$ time python -c 'print ("".join(map(lambda x: ("[" + str(x) + "]: u\n"), range(1,10000 * 16))) + "[0] " * 10000)' | ./hoedown >/dev/null
real	0m8.778s
user	0m8.739s
sys	0m0.029s

[markand]

Are you really saying that 8 seconds is too big for 160000 links? And as a note, if you really have a document as big I'm not sure if I would read it 😁

[mity]

Well, yes, it is slow. Maybe I should elaborate more.

Parsing few MB in seconds is slow on a modern computer.

Much worse, the time grows in much worse fashion (I guess quadratic) then linear here. On my quite fast machine I get these results, with this slightly more generalized example for different N:

$ time python -c 'N=16; print ("".join(map(lambda x: ("[" + str(x) + "]: u\n"), range(1,N*10000))) + "[0] " * N*10000)' | ./hoedown >/dev/null

And, for comparison, I added times it takes MD4C parser.

N	input size	hoedown	md4c
8	1.2MB	5.859s	0.125s
16	2.5MB	23.531s	0.203s
24	3.8MB	55.000s	0.281s

And no, it's not about whether you read the document or not.

If an application can read the document from any (read: untrusted) source, this document can be used as a DOS attack, even if the application limits document size to e.g. something like 10MB, consuming CPU cycles of your server for long minutes per single request.