...and just because this might be of interest:

I've had a look into the question I've placed in bold above...

I created a dataset of 5 million numbers each with a random number of digits (between 1 and 10 digits). 10 numbers per line, each separated by a space.

Then I used time to time 10 runs of an awk program which used [0-9][0-9][0-9][0-9]+ and then 10 runs with [0-9][0-9][0-9][0-9][0-9]*.

Since it was being run on my Linux desktop PC, I used chrt and set the scheduling policy to SCHED_FIFO with a priority of 99 which as far as I know gives the process the highest priority possible. The commands were:

Code:

chrt -f 99 time -f "\n***\nSecs: %e \nCPU: %P \nContext Switches: %c \nWaits: %w"
awk 'BEGIN { regex = "[0-9][0-9][0-9][0-9]+"; } { line = $0; gsub(regex, "x", line); }'
< NumsData5MillionNums >> ResRegex4 2>&1

chrt -f 99 time -f "\n***\nSecs: %e \nCPU: %P \nContext Switches: %c \nWaits: %w"
awk 'BEGIN { regex = "[0-9][0-9][0-9][0-9][0-9]*"; } { line = $0; gsub(regex, "x", line); }'
< NumsData5MillionNums >> ResRegex5 2>&1

I don't think the results can be considered as particularly scientific... But they were fairly consistent. BTW as expected each run had 0 context switches and 1 wait.

In fact the results were so close that I think that the Awk interpreter was probably running the same code in both cases, after all the 2 regexes [0-9][0-9][0-9][0-9]+ and [0-9][0-9][0-9][0-9][0-9]* are logically interchangeable.

I sorted the times and discarded the 3 fastest and 3 slowest times of the 10 runs, leaving me with:

Code:

ResRegex4 - regex = "[0-9][0-9][0-9][0-9]+" :

Secs: 7.16
Secs: 7.21
Secs: 7.27
Secs: 7.28

Mean: 7.23
Median: 7.24

ResRegex5 - regex = "[0-9][0-9][0-9][0-9][0-9]*" :

Secs: 7.14
Secs: 7.17
Secs: 7.18
Secs: 7.25

Mean: 7.185
Median: 7.175

Full output of "[0-9][0-9][0-9][0-9]+" is here: http://pastebin.com/VqC5dbna

Full output of "[0-9][0-9][0-9][0-9][0-9]*" is here: http://pastebin.com/U6rpULd6

The C code to create the data file of 5 million numbers, each 1-10 digits in length, and with 10 numbers on each line is here: http://pastebin.com/6vG9WQwj