@MrSwiss : you are right. It was nevertheless a real challenge. 2x slower could have been an issue, 8x slower definitely a pitfall. The tool still retains some interest, but this allows to better measure the limits. When the length of the strings exceeds a certain threshold and there is no implicit ...

After accessing my computer again, .. My tests show that compared to the algorithm of Dodicat, Lzle is 8 to 9 times slower and especially that the need memory is much more important, in unacceptable proportions. I conclude that the list-based tool can not compete with an array in the proposed case, ...

Speed results can be very different on a random dataset depending on you fall in a slowdown case or not on the related algo (quicksort, bubble, and so on...). Speed can be also related to cache, processor, os... I can see at least 2 different approach : going on a fast algo and trying to reduce or l...

Considering JK explanations and extrapolating to lzle it is now obvious for me than the case independant feature should be handled into the hashtag (and not before it) in this context but I m not sure if it would be worth or not to add such feature to the tool (one more test and a small but global s...

2-3 sec with a len 10-20 on my side. Using len 60-90 could be 4 times slower or above. to be tested using different optimization options. whenever your solution is using disk cache or not, sounds interesting for optimization. I ll do a few tests also on a crypted disk. I ll do also some new tests on...

On my side : i5 2410m w64, battery mode. Lzle.bi (FB) : 2 sec / 3 sec depending on the dataset.

"more code, more trouble, less speed" In general, I'd say yes . But in the case of a sort, the problem is much more complex. Certainly, a code "bigger" will consume cache, but if the object of this code is precisely to limit cache consumption by the data, then this choice become...

I remain convinced of the relevance of lzle in this case (about 1000000 entries): it is a quasi standard feature of the tool. The reuse of the memory addresses makes the performances much less sensitive to the structure of the data to be sorted and the successive sorts and the tree is adapted to ave...

Yes, instead of deleting outside, a redim inside a parallel scope (triggered by a parameter passed to the function) will indicate to the compiler the free space. It is difficult to imagine Dim doing the job just one way. lzae envisaged evolutions : support for indices on SetCursor(indice, (...) as i...

Sounds like a "quirk" !-) #Include Once "D:\Basic\lzae_04.bi" Function returnPtr(ByRef aext As ArrayExtension, nB as Integer=2) as integer Static outarr() as double '= {10.1,20.2,30.3} Redim Preserve outarr(nB) outarr(0)=10.1 : outarr(1)=20.2 : outarr(2)+=30.3 : aSet(outarr() , a...

Even though at the technical level in the Globals section, Static means "persistent" (it is rather a scope of duration than mapping). If the array is not copied, it ceases to exist at the end of the returnPtr function. Starting with the principle of creating the array in the function, usin...

something like this : #Include Once "D:\Basic\lzae_04.bi" Function returnPtr(ByRef aext As ArrayExtension) as integer Static outarr(...) as double = {10.1,20.2,30.3} aSet(outarr() , aext) Return 1 End Function Dim i As Integer Dim aext As ArrayExtension returnPtr(aext) For i=1 to aext.NbSt...

What really is an array in FB ? It could be summarized as two data structures and a hidden algorithm. The second structure contains all the elements of the array following each other and the first data structure (called the descriptor) contains technical access information (address of the first and ...

@fxm. Certes, certes. mais cela montre que quelqu'un est capable de faire des posts valant les miens :-)