Vastly simplify C version about py-lmdb

[jstasiak]

I've been wondering, what's the difference between C and CFFI versions in terms of performance?

[dw]

Hi Jakub!

It depends on the particular operation, but typically 2-3x for tight loops doing get()/iteration within a transaction, but perhaps 10x or worse if you're creating many transactions/cursors/iterators at a fast rate.

Thanks for asking – because it seems the cffi build is currently broken, due to new version of cffi. There is a script, examples/dirtybench.py that gives rough performance guides, so once cffi is fixed you can see for yourself

[dw]

Hi Jakub,

I coaxed it to run.

CFFI: https://gist.github.com/dw/4606e31f3e69163097db/raw/adedde4d9cb32fde480b80e34d6313b52d46d8fe/gistfile1.txt

C: https://gist.github.com/dw/b32524ffcbcce574ab8c/raw/85f19cfc6df3e68b4a0bc1ebc47f56158c5e5913/gistfile1.txt

[dw]

Freelists probably not being done, so closing this.

[dw]

Freelists probably not being done, so closing this.