Numeric Maps

[mk-soft]

Hello Fred,

Please publish the internal existing Numeric Map functions as purebasic functions.
NewNumericMap, etc

Thanks

[jacdelad]

Was requested numerous times and I totally give it a +1!

[skywalk]

Is there a significant speed improvement with map access using integer keys?

[ChrisR]

+1, It would be really nice to have it natively to avoid AddMapElement(Map(), Str(Key)), FindMapElement(Map(), Str(Key))
and a second +1 if the function is already there

[Tenaja]

Time flies; I asked about this ten years ago. There are some solutions in this thread:
https://www.purebasic.fr/english/viewto ... 38#p440238

[mk-soft]

+1, It would be really nice to have it natively to avoid AddMapElement(Map(), Str(Key)), FindMapElement(Map(), Str(Key))
and a second +1 if the function is already there

Would have to be nicely implemented by Fred
Link: https://www.purebasic.fr/english/viewtopic.php?t=74214

[ChrisR]

Thanks for the link

Messing with internal code isn't really a trick and tips as it can be broken anytime. If it's not released, it's because it's ready for

Note to self, don't unwrap Fred's xmas presents

For Christmas it's too late but we have until January 31st to wish to all a happy new year
or else it's good to keep dreaming that one day it will arrive

[Quin]

+1. This would save me from calling Str() in so many places and definitely improve performance.

[Caronte3D]

+1 For sure

[idle]

Thanks for the link

Messing with internal code isn't really a trick and tips as it can be broken anytime. If it's not released, it's because it's ready for

Note to self, don't unwrap Fred's xmas presents

For Christmas it's too late but we have until January 31st to wish to all a happy new year
or else it's good to keep dreaming that one day it will arrive

Use Squint3 it's a bit faster than a map, grows dynamically and is faster to exit if a key doesn't exist
https://www.purebasic.fr/english/viewtopic.php?t=79453

[ChrisR]

Use Squint3 it's a bit faster than a map, grows dynamically and is faster to exit if a key doesn't exist
https://www.purebasic.fr/english/viewtopic.php?t=79453

Promise the next time I need a map with a numeric or alphanumeric key and a bit of performance, I'll try Squint3
I had already seen your great Module (like other great code from you. Looking at them makes me feel like a noob) but I haven't used it yet.
Code with embedded C or ASM has always scared me a bit, because I don't know anything about them, I don't know what's being done and I can't control the code if I need to. Even if I'm confident

[AZJIO]

The key is not just a string, but the binary data of the string. Perhaps another good way is to pass a pointer to a string, then it would be possible to add another parameter - the length of the string StringByteLength(). If the length is not specified, then it is determined by 00, and if a pointer to a number is used, then SizeOf can be specified. In any case, the data will no longer have a special type.

[idle]

Use Squint3 it's a bit faster than a map, grows dynamically and is faster to exit if a key doesn't exist
https://www.purebasic.fr/english/viewtopic.php?t=79453

Promise the next time I need a map with a numeric or alphanumeric key and a bit of performance, I'll try Squint3
I had already seen your great Module (like other great code from you. Looking at them makes me feel like a noob) but I haven't used it yet.
Code with embedded C or ASM has always scared me a bit, because I don't know anything about them, I don't know what's being done and I can't control the code if I need to. Even if I'm confident

truth be told I can barely control the code and I wrote it.

The key is not just a string, but the binary data of the string. Perhaps another good way is to pass a pointer to a string, then it would be possible to add another parameter - the length of the string StringByteLength(). If the length is not specified, then it is determined by 00, and if a pointer to a number is used, then SizeOf can be specified. In any case, the data will no longer have a special type.

squints Trie functions can be string keys or binary keys using the binary functions (but you don't mix binary and string keys in same trie)
string keys accepted are unicode(UCS2), utf8 or ascii null terminated strings (all keys are mapped to utf8 to save space)
keys are sorted on input into the Trie so you can query them by prefix, to get a sorted output in a the callback

binary keys can be anything you want up to #SQUINT_MAX_KEY = 1024 just set the size and a value, they are currently only on the root and shouldn't be used together with strings or numeric keys in the same Trie.

The map function is numeric or binary and is numerically sorted
it accepts integer keys up to sizeof(integer) and a value
or a binary key up to #SQUINT_MAX_KEY = 1024 when you set the hash flag, the binary key is then hashed
it behaves a bit like a HAT (a hash collision could be possible but very unlikely)
don't mix numeric and binary hashed keys together if you want to dump the trie as you won't be able to tell what it is

When I say strings and integer keys can both be used together in the same trie, it's only valid for numbers up to 60 bits because the numeric keys are on the roots child[0] and strings wont share that root. but you can just create another trie
an empty trie only takes up 1328 bytes which can be reduced to 304 bytes when I get around to fixing it.

maybe fred will put us out of the misery and implement the numeric map.

[moricode]

Is there a significant speed improvement with map access using integer keys?

definitely faster and less conflict as integer key doesn't need to hash at all , it just need to divide the slot number and store in table only .

example :
myintmap(12345) = 9899

bucket slot size = 512
so the key = 12345

table position = 12345/512 =24 (truncated )
store the value 9899 in table position 24

conflict only if another ==> key/512 = 24 than you need chaining in table for the double entry value

over all , it is more efficient and fast , less conflict .

make one for yourself for fun

[Fred]

It's not published because I want a seemless integration (not adding AddNumericMapElement() for example but still use AddMapElement() with a numeric param) but it needs some compiler tweaks.