Tuesday, 12 June 2012

Undocumented feature - interrupt - part three

Continuing from part two where we looked at "0" and "1"...

2 - Extract filename

This seems to extract the filename part of a full path, removing the directory structure.  Here's some example code for you...

filename = $result

In this case "filename" will be populated with "text.txt".  This does seem to be a simple case of string manipulation, but if you support both Windows and Unix filesystems (which the company I work for do, and I expect a lot of people will if they are doing any web development) then maybe it's not so simple...

filename = "C:\temp\test.txt"
pos = $rscan(filename,"\";for Windows
if ( pos < 1 )
  pos = $rscan(filename,"/";for Unix
filename = filename[pos+1]

So this requires a numeric local variable and a few extra lines of code to achieve.  I tested for 2,000,000 iterations...

  • interrupt = 00:09.35, 00:09.30, 00:09.33 (over 9 seconds)
  • $rscan (Windows first) = 00:14.72, 00:14.75, 00:14.66 (under 15 seconds)
  • $rscan (Unix first) = 00:16.38, 00:16.41, 00:16.31 (over 16 seconds)

Unfortunately $rscan is a pretty slow function, but as they could be multiple folder delimiters in the string, I don't see how it can be avoided.  You have to choose in your code whether to put Windows or Unix first, unless you have a global setting somewhere which tells you which the application is running on, in which case you could limit this to a single $rscan.  

Whilst interrupt does perform better in this case, I would say that it is more than likely negligible, given the number of iterations.  Therefore, I'd choose to use the alternative, as I know that I can rely on it working in future.

3 - Extract file path

This seems to extract the file path part of a full path, removing the filename.  Here's an example...

filepath = $result

In this case "filepath" will be populated with "C:\temp\".  The alternative to this is therefore going to consist of a similar bit of string manipulation...

filepath = "C:\temp\test.txt"
pos = $rscan(filepath,"\";for Windows
if ( pos < 1 )
  pos = $rscan(filepath,"/";for Unix
filepath = filepath[1:pos]

So there's no need to measure the performance of this, the different will be very similar.

Summary: Numeric codes "2" and "3" do seem to be useful and perform better than the alternatives.  However, the performance difference is probably not enough to consider using the interrupt command, as it will be less maintainable and risky moving forwards.  I'll continue in the next part.

No comments:

Post a Comment