Hi, I'm building trigrams for text. And not sure how to optimise it.
1. Nim won't allow to use `addr` and suggest to use `unsafe_addr` instead,
why?
2. The code fails with `unsafe_addr`
3. It seems that the whole idea to store StringSlice in HashTable is wrong.
The trigram string should be stored in HashTable, and Slice should be used only
to get the code from the table, and not be stored there. But `Table[string,
int]` won't allow to be queried by `table[key_of_string_slice_type]`.
import std/[tables, hashes, strutils]
type SSlice = object
s: ptr string
i, l: int
proc sslice*(s: ptr string, i: int, l: int): SSlice =
SSlice(s: s, i: i, l: l)
proc `$`*(s: SSlice): string =
for j in 0..(s.l - 1): result.add s.s[s.i + j]
proc `==`(a, b: SSlice): bool =
if a.l != b.l: return false
for j in 0.int..(a.l - 1):
if a.s[a.i + j] != b.s[b.i + j]: return false
true
proc hash*(s: SSlice): Hash =
result = 0
for j in 0..(s.l - 1):
result = result !& s.s[s.i + j].hash
result = !$result
var codes: Table[SSlice, uint16]
template encode_trigram*(s: SSlice): uint16 =
var r: uint16
with_value(codes, s, value):
r = value[]
do:
codes[s] = codes.len.uint16
r = codes[s]
r
proc to_trigrams*(text: string): seq[uint16] =
if text.len < 3: return text.align_left(3).to_trigrams
for i in 0..(text.len - 3):
result.add sslice(unsafe_addr text, i, 3).encode_trigram
let last_two = text[^2..^1] & " "
result.add sslice(unsafe_addr last_two, 0, 3).encode_trigram
let last_one = text[^1..^1] & " "
result.add sslice(unsafe_addr last_one, 0, 3).encode_trigram
echo "some text".to_trigrams
echo "some".to_trigrams
Run
P.S.
4. Also, not sure why this code fails?
type SSlice = object
s: string
i, l: int
template sslice*(s: string, i: int, l: int): SSlice =
SSlice(s: s, i: i, l: l)
echo sslice("some", 0, 3)
Run
.nim(10, 13) Error: identifier expected, but found '"some"'
Run
