PureBasic 6.40 alpha 1 is ready, surprise inside !

[akee]

Thanks Fred.

[threedslider]

MessageRequester() - 267 KB.
Empty source file - 265 KB.

Let's compare that with PureBasic 5.73:

MessageRequester() - 4.50 KB.
Empty source file - 2 KB.

That's why I still use 5.73 for my private projects. The old PureBasics were so lean and clean.

[DeanH]

The change to string handling sounds similar to the old BSTR string type. Much faster to read records from a large SQLite database. Question: how is the length stored? Long? Quad? Another way? Does it limit the maximum length of a string?

[PrincieD]

Wow! that sounds awesome Fred, so instead of reading a string for null termination - the length is already cached?

[Little John]

Wow! that sounds awesome Fred, so instead of reading a string for null termination - the length is already cached?

Yes, that is the case:

All strings are now prefixed by their length (except the strings in DataSection). [...]
That means than all string operation will have instant length information resulting in faster operation as iterating byte by byte on the string to find the null terminated char is no more needed.

It's the way e.g. Turbo Basic did it already about 35 years ago.

[Fred]

The change to string handling sounds similar to the old BSTR string type. Much faster to read records from a large SQLite database. Question: how is the length stored? Long? Quad? Another way? Does it limit the maximum length of a string?

It's an integer minus 1 bit (31 bit on x86, 63 bit on x64).

[Matheos]

The change to string handling sounds similar to the old BSTR string type. Much faster to read records from a large SQLite database. Question: how is the length stored? Long? Quad? Another way? Does it limit the maximum length of a string?

It's an integer minus 1 bit (31 bit on x86, 63 bit on x64).

I appreciate you said in the opening post that the terminating null is still retained, but could this possibly lead eventually to an ability to store x'00 in string values?

I support an application that provides independent string functions, written primarily to do this, although with performance objectives in mind also, so I'm keen to experiment further when I have chance and see if we can begin to take advantage of this change.

[Fred]

Theorically, you can put 0 in it. But it won't be officially supported for now (could evolve later but will be per command decision)

[skywalk]

Fred - did you consider making another string type?
Ex. Cstring for api, FastString for internal?

Or is the plan to have new string support all api calls with null terminated strings?

[Fred]

There is no plan to have an additional string for API call, as they are always null terminated. The only case when it needs an adjustment is when you use a string for output buffer using Space().

[HeX0R]

A probably easier method to recalc the length of the string, without any fancy additional procedures:

Code: Select all

PureBasicPath$ = Space(#MAX_PATH)
GetModuleFileName_(GetModuleHandle_(#Null$), @PureBasicPath$, #MAX_PATH)
Debug Len(PureBasicPath$)
PureBasicPath$ = PeekS(@PureBasicPath$)
Debug Len(PureBasicPath$)

Edit
oh god, my bad, Fred even mentioned that in the initial post

Using Win32 API with Space() for example will require an extra PeekS().

[ChrisR]

Or by using fryquez ResetStringLength procedure/macro: viewtopic.php?p=650721#p650721

Code: Select all

CompilerIf #PB_Compiler_Version >= 640
  Declare ResetStringLength(*s.Character)
  
  Procedure ResetStringLength(*s.Character)
    Protected *p.Integer = *s -SizeOf(Integer)
    *p\i = 0
    While *s\c
      *p\i + 1
      *s + SizeOf(Character)
    Wend
  EndProcedure
CompilerElse
  Macro ResetStringLength(s)
    ;
  EndMacro
CompilerEndIf

CompilerIf Not #PB_Compiler_Debugger
  OpenConsole("ResetStringLength")
  Start = ElapsedMilliseconds()
  For l = 0 To 5000000
    PureBasicPath$ = Space(#MAX_PATH)
    GetModuleFileName_(GetModuleHandle_(#Null$), @PureBasicPath$, #MAX_PATH)
    PureBasicPath$ = PeekS(@PureBasicPath$)
  Next
  PrintN("PeekS Len(" + Str(Len(PureBasicPath$)) + "): " + Str(ElapsedMilliseconds() - Start) + " ms")
  
  Start = ElapsedMilliseconds()
  For l = 0 To 5000000
    PureBasicPath$ = Space(#MAX_PATH)
    GetModuleFileName_(GetModuleHandle_(#Null$), @PureBasicPath$, #MAX_PATH)
    ResetStringLength(@PureBasicPath$)
  Next
  PrintN("ResetStringLength Len(" + Str(Len(PureBasicPath$)) + "): " + Str(ElapsedMilliseconds() - Start) + " ms")
  
  Delay(10000)
CompilerEndIf

;PeekS Len(61): 517 ms
;ResetStringLength Len(61): 350 ms

[freak]

Or by using fryquez ResetStringLength procedure/macro: viewtopic.php?p=650721#p650721

Code: Select all

CompilerIf #PB_Compiler_Version >= 640
  Declare ResetStringLength(*s.Character)
  
  Procedure ResetStringLength(*s.Character)
    Protected *p.Integer = *s -SizeOf(Integer)
    *p\i = 0
    While *s\c
      *p\i + 1
      *s + SizeOf(Character)
    Wend
  EndProcedure
CompilerElse
  Macro ResetStringLength(s)
    ;
  EndMacro
CompilerEndIf

CompilerIf Not #PB_Compiler_Debugger
  OpenConsole("ResetStringLength")
  Start = ElapsedMilliseconds()
  For l = 0 To 5000000
    PureBasicPath$ = Space(#MAX_PATH)
    GetModuleFileName_(GetModuleHandle_(#Null$), @PureBasicPath$, #MAX_PATH)
    PureBasicPath$ = PeekS(@PureBasicPath$)
  Next
  PrintN("PeekS Len(" + Str(Len(PureBasicPath$)) + "): " + Str(ElapsedMilliseconds() - Start) + " ms")
  
  Start = ElapsedMilliseconds()
  For l = 0 To 5000000
    PureBasicPath$ = Space(#MAX_PATH)
    GetModuleFileName_(GetModuleHandle_(#Null$), @PureBasicPath$, #MAX_PATH)
    ResetStringLength(@PureBasicPath$)
  Next
  PrintN("ResetStringLength Len(" + Str(Len(PureBasicPath$)) + "): " + Str(ElapsedMilliseconds() - Start) + " ms")
  
  Delay(10000)
CompilerEndIf

;PeekS Len(61): 517 ms
;ResetStringLength Len(61): 350 ms

Just don't act surprised when that code breaks in the future because you are clearly messing with internals here. There is such a thing as over-optimizing.

[mk-soft]

The test on NIL (null pointer) is missing

[diceman]

Basically if you patch a string by putting a zero in it, the Len() function will be wrong, and the concat functions will fail. You will need to use PeekS() for this.

Please, for really stupid people like me, can you explain in layman's terms what that actually means?
Kinda afraid to install the patch, as I am doing a lot of String Operations, and seeing anything break that worked before in my 65k lines-project, would just cause unneccessary stress to me, lol, thanks.

- What does "patching a string by putting a zero in it" even mean?
- What are concat functions?
- An snippet of a thing that worked before, and after the patch has to be done with PeekS (never used that one either) would be great.