Page 1 of 1
Some test data for strings
Posted: Wed Feb 19, 2020 5:29 pm
by mainhatten
Hi,
as I am working on vfp GetWordNum/GetWordCount (perhaps to be reused or -visited with alines()), test data was needed.
I remembered an old article of Steve Black,
https://github.com/StevenBlack/sbc.next ... n-VFP.html
and searched for the raw text, which (of course) had moved, but I found some Tolstoy texts elsewhere.
4 records (War&Peace, Anna Karenina, Resurrection and Kreutzer Sonata) tucked away in memo fields of vfp table and zipped as attachment.
Hope it is sufficient enough as basis for "long string work" - some web sites deliver similar amounts to War&Peace...
Perhaps others need long texts as basis as well - if so, use without FileToStr() or LLF calls.
regards
thomas
Some test data for strings
Posted: Fri Feb 21, 2020 10:16 am
by Karl-Heinz
Hi Thomas,
i´ve found the 3.2 MB "war-and-peace.txt" and made some tests to see how fast some of the X# string and array funcs are. And they are very fast ! Of course the slowest way is to use AAdd(), but keep in mind that there are 65007 elements that must be added step by step.
The required references are Xsharp.Core.dll and XSharp.rt.dll
https://github.com/fluentpython/example ... -peace.txt
Code: Select all
FUNCTION ReadWarAndPeaceTxtFile()AS VOID
LOCAL t1, t2 AS FLOAT
LOCAL cValue, cFile AS STRING
LOCAL h AS PTR
LOCAL i, dwLines, dwHits AS DWORD
LOCAL a, b, aStrings AS ARRAY
// https://github.com/fluentpython/example-code/blob/master/attic/sequences/war-and-peace.txt
// ------------------------------------------------
cFile := "D:testvfpwar-and-peace.txt"
// ------------------------------------------------
t1 := Seconds()
cValue := MemoRead ( cFile )
t2 := Seconds()
? "MemoRead() seconds:" , t2-t1 // 0.03 secs
t1 := Seconds()
dwLines := MLCount( cValue ) // 65007 lines
t2 := Seconds()
? "MLCount() seconds:" , t2-t1 , "number of lines:" , dwLines // 0.03 secs
? "Last line:" , MemoLine( cValue , , dwLines )
aStrings := ArrayNew ( 5 )
aStrings[1] := "Russia" // 775 occurences
aStrings[2] := "Anna" // 293 occurences
aStrings[3] := "Czar" // 4 occurences
aStrings[4] := "windows" // 23 occurences
aStrings[5] := "Pentium" // 0 occurences
FOR i := 1 UPTO ALen ( aStrings )
t1 := Seconds()
dwHits := Occurs( aStrings[i], cValue )
t2 := Seconds()
? 'Occurs ("' + aStrings[i] + '" , cValue) ' , "" , "Hits:" , dwHits , "" , "seconds: " , t2 -t1
NEXT
?
IF ( h := FOpen( cFile )) != F_ERROR
cValue := ""
t1 := Seconds()
FRead( h, @cValue, 4000000 )
t2 := Seconds()
? "FRead() seconds:" ,t2 -t1 // 0.03 secs
? "content of last line:" , MemoLine( cValue , , dwLines )
// --------------------
a := ArrayNew ( dwLines )
FRewind(h)
t1 := Seconds()
FOR i:= 1 UPTO dwLines
a [i] := FGetS ( h )
NEXT
t2 := Seconds()
? "ArrayNew() seconds:" , t2-t1 , "Elements:" , ALen ( a ) // 0,33 secs
? "content of last line:" , ATail ( a )
// -----------------
?
b := {}
FRewind ( h )
t1 := Seconds()
DO WHILE ! FEof ( h )
AAdd ( b, FGetS ( h ) )
ENDDO
t2 := Seconds()
? "AAdd() seconds:" , t2-t1, "Elements:" , ALen ( b ) // 0,53 secs
? "content of last line:" , ATail(b)
?
FClose( h )
ENDIF
RETURN
regards
Karl-Heinz
Some test data for strings
Posted: Fri Feb 21, 2020 10:40 am
by SHirsch
Hi Karl-Heinz,
I tested you prog.
My times differs from yours:
ArrayNew() 1.33 sec
Aadd() 1,53 sec
Than I checked build in .Net method:
Code: Select all
t1 := Seconds()
VAR x := System.IO.File.ReadAllLines(cFile)
t2 := Seconds()
? "ReadAllLines seconds:" , t2-t1, "Elements:" , x:Length // 0,02 secs
? "content of last line:" , x[x:Length]
- KH-Test.jpg (66.78 KiB) Viewed 2401 times
Conclusion: Refactorings are seldom wasted time.
Stefan
Some test data for strings
Posted: Fri Feb 21, 2020 12:11 pm
by Karl-Heinz
Hi Stefan
i added your ReadAllLines() test and the time elapsed is 0,02 secs. About your ArrayNew() and AAdd() results: I cleaned my eyeglasses but my results are still 0,35 and 0,56 secs - always about 1 sec faster than your results. Ok, let´s wait what others report.
regards
Karl-Heinz
Some test data for strings
Posted: Fri Feb 21, 2020 12:38 pm
by SHirsch
Hi Karl-Heinz,
haha, I took an old test app referencing Vulcan dlls. Changed this to XSharp an now I can copy your results (0,31 and 0,52 and 0,02).
Stefan
Some test data for strings
Posted: Fri Feb 21, 2020 1:41 pm
by Karl-Heinz
Hi Stefan,
Indeed, if the VN runtime is used instead, the ArrayNew() and AAdd() results are similar to your first post !
regards
Karl-Heinz
Some test data for strings
Posted: Fri Feb 21, 2020 2:18 pm
by Chris
Guys, I think it's also worth comparing the memory consumption off the two versions (X#/vulcan runtime)
Some test data for strings
Posted: Fri Feb 21, 2020 8:24 pm
by Karl-Heinz
Hi Chris,
i found this on GitHub :
https://github.com/X-Sharp/XSharpPublic ... /Start.prg
and modified it a bit. Compared to X#, VN needs a lot more memory.
VULCAN
Memory for 65007 element ARRAY 2.600.332 bytes
Memory for 65007 element ARRAY after assigning 65007 values 2.600.332 bytes
Memory for 65007 element ARRAY after assigning 65007 values with AAdd 2.602.480 bytes
Memory for 1000000 element ARRAY 40.000.056 bytes
Memory for 1000000 element ARRAY after assigning 1000000 values 40.000.056 bytes
Memory for 1000000 element ARRAY after assigning 1000000 values with AAdd 40.195.216 bytes
X#
Memory for 65007 element ARRAY 1.040.192 bytes
Memory for 65007 element ARRAY after assigning 65007 values 1.040.192 bytes
Memory for 65007 element ARRAY after assigning 65007 values with AAdd 1.048.636 bytes
Memory for 1000000 element ARRAY 16.000.080 bytes
Memory for 1000000 element ARRAY after assigning 1000000 values 16.000.080 bytes
Memory for 1000000 element ARRAY after assigning 1000000 values with AAdd 16.777.276 bytes
Anyway, VN is history
regards
Karl-Heinz
Some test data for strings
Posted: Fri Feb 21, 2020 11:56 pm
by Chris
Yeap
.
We had been discussing this internally years ago, how to implement USUALs, ARRAYs, FLOATs etc in X#, so that they are much more efficient than in Vulcan. It payed off
Some test data for strings
Posted: Wed Feb 26, 2020 6:42 pm
by mainhatten
SHirsch wrote:haha, I took an old test app referencing Vulcan dlls. Changed this to XSharp an now I can copy your results (0,31 and 0,52 and 0,02).
Great - comparing and finding the better implementation/setting to use is one of the things better done with same data.
I read up on vfp9 Eula and Redistrib.txt, and the data files given with vfp examples (Northwind, TastradeData and Data_for_samples) can be deployed. All of these are rehashes of very similar data pool, but needed if one tries to port vfp samples (guite old, must look if a good example for cursoradapter, buffering, conflict update is among them), but could/should serve as data basis for any examples, as no problem with data protection laws exist there.
Also, if an example or speed test needs huge amount of data, it is easier to write a few lines of xBase appending slightly modified data to itself, which one can run on own machine without DL of a few hundred MB.
Did not want to spam vfp/MS data here, but as there seems to be no special Clipper/VO/VN testdata, building on the old MS stuff is IMHO better than creating always new data. Any interest in having common play data ?
If so, I'll install (typically not installed here on dev VM), zip and attach to a post. All three only a few MB IIRC, but it is nice to have tables set up where relational links make some sense and have that structure filled with data if I want to show something from a SQL syntax to GUI example.
As long as path to directory name can be pointed to in example, easy to follow any example running on same DB on own hardware.
regards
thomas