Hacker News Reader

I lowkey wonder how much memory is consumed, on average, by the extra string info in map-heavy languages like JS and Python, and how much CPU time is spent on strings there. While C doesn't need too many strings and mostly throws around numbers, a Node.js app may easily be dealing with thousands of them, not even counting HTML and templates.

(And now I also wonder if any compilers and VMs precompute hashes for map and object keys.)

View on HN

Slightly related: Lua will intern every string ever created, that is two strings are equal if and only if their pointers (as represented internally) are identical. This results in slow string creation, but very fast equality checking and lookups, as it converts the problem to comparing integers.

> (And now I also wonder if any compilers and VMs precompute hashes for map and object keys.)

I don't think i've ever come across one that does. But then, in most situations where they could, they can do something even more efficient anyway. Fast JS compilers replace an object's hashtable with a dense array of fields, accessed by index. Lots of language implementations do method lookup in a vtable, rather than a dictionary.

Reply on news.ycombinator.com