timtadh / data-structures Goto Github PK

Hi, I used the hashtable from your library. When I use the hashtable and often got hit miss, the library generates a lot of stack. It turns out to be an expensive operation.
Could we remove it and just use error wrapping instead?

Your readme states that you provide "generic types to allow the user to swap out various data structures transparently. The interfaces provide operation for adding, removing, retrieving objects from collections as well as iterating over the collection using functional iterators."

How does this work? For example:

var myKey int16 = 1
table := NewHashTable(10)
var key types.Int16
// XXX: key = myKey
var value interface{} = myData
var err error
value, err = table.Put(key, value)

I don't understand how to assign myKey to the types.Int16 key.

Thank you.

Hi:
this is the case

package main

import (
	"github.com/timtadh/data-structures/trie"
)

func main() {
	t := trie.New()
	for _, _, next := t.Iterate()(); next != nil; _, _, next = next() {

	}
}

golang version: go version go1.11.4 linux/amd64
when i'm using go run, it occurs

panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x0 pc=0x48ac3b]

goroutine 1 [running]:
github.com/timtadh/data-structures/trie.(*TST).Iterate.func1(0xc000067788, 0xc000090020, 0x0, 0x0, 0x0)
	/home/tcp9/go/src/github.com/timtadh/data-structures/trie/tst.go:215 +0x3b
main.main()
	/home/tcp9/go/src/test/test/main.go:9 +0x62

https://github.com/timtadh/data-structures/blob/master/trie/tst.go#L173

This line panics if the prefix being searched for is longer than the capacity of the key being checked, e.g. this panics:

	tst := trie.New()
	tst.Put([]byte("abc"), nil)
	_, _, _ = tst.PrefixFind([]byte("abcxxxxxxxxxxxxxxxxxxxx"))()

panic: runtime error: slice bounds out of range [:23] with capacity 8
...
github.com/timtadh/data-structures/trie.(*TST).PrefixFind(0xc000092760?, {0xc000092749, 0x17, 0x3?})
	/go/pkg/mod/github.com/timtadh/[email protected]/trie/tst.go:173 +0x297
...

Since the slice operation succeeds if the index is larger than the length but smaller than the capacity this doesn't always occur for prefixes that are slightly longer than the found leaf node.

Several places in your code, you open /dev/urandom directly, and use that to feed into math/rand.Seed(). This breaks Windows tests, as /dev/urandom is Linux only.

Consider switching to crypto/rand.Read() as seed source instead, or just using current timestamp to seed with, that's universal.

Additionally in sortedset.go (init) you should not apply seeding. It's not really your library's responsibility to do so. For testing, modules, yes, but otherwise actually it's the programmers responsibility (perhaps they want to keep it unseeded for reproducable testing).

As it currently stands. It seems I can only use "types.Int" as keys in the bptree is there a way we can use float64 instead?