Possible bug in split() in awk95

[d0vgan]

Hello, everybody.

Does anybody use awk95 by Brian W. Kernighan? I think, I found a bug in realization of function split(), but I am not sure.
Here is a sample code where you can see it:

BEGIN {
time = "10:30:00,000"
split( time, a, ":" )
print a[3] # prints 00,000 - it's OK
i = split ( a[3], a, "," ) # I want to split "00,000" by ","
print i # prints 2 - it's OK
print "a[1] = " a[1] # prints garbage, must be 00
print "a[2] = " a[2] # prints garbage, must be 000
}

Regards,
Vitaliy.

 

[feherke]

Hi

Well, I have [tt]awk95[/tt], but on Windows I prefer to use [tt]gawk[/tt]...

Indeed, the error you described appears. I would not use the same array [tt]a[/tt] for both source and destination. Then that problem does not appear.

Better install CygWin and use a [tt]gawk[/tt], the GNU implementation of [tt]awk[/tt] compiled into native Windows executable.

Feherke.

http://rootshell.be/~feherke/

 

[PHV]

The problem is here:
i = split ( [!]a[/!][3], [!]a[/!], "," )

I confirm that gawk handles this situation pretty well, but legacy awk (even nawk) don't.

Hope This Helps, PH.
FAQ219-2884
FAQ181-2886

 

[feherke]

Hi

Pity. [tt]nawk[/tt] looked far better in my eye. Even [tt]mawk[/tt], which implements only the base functionalities, can handle this situation.

Feherke.

http://rootshell.be/~feherke/

 

[PHV]

Anyway, I insist on the poor programming way followed by the OP ...
I personally don't see any bug here.

 

[d0vgan]

PHV, you wrote:
> I personally don't see any bug here.

Well, if such action as split(a[3], a, ",") is incorrect, then there must be some warning from AWK, for example: "WARNING: you can not split an array element into the same array".
But awk95 silently returns correct number of subfields and corrupted data in the subfields. It is a data loss which seems to be a real bug.

Talking about awk95 and mawk, I think that awk95 has a big advantage: though mawk is faster, mawk does not support 64-bit integers. awk95 automatically switches from 32-bit integer to 64-bit integer when it's neccesary - and it is very important for calculations with large integer numbers or working with databases with high indexes.