Exiftool Test

[AlainP]

I just made a quick test with exiftool, in consumer beta .47

 

Here is the time it took (approx.) to open a RAW CR2 file (28 Mb)

 

Included version 10.1.6.0 :  19 seconds

New version 10.3.7.0 :       19 seconds

Exiftool disabled:               11 seconds

 

I don't have the speediest computer but AP runs very well on it, and I think that 8 seconds just to read the metadata from the RAW file is not normal. So this is a big part of the time it takes to open a RAW file in AP. This operation can probably be optimized and speed up AP.

 

 

[lenmerkel]

I don't have the speediest computer but AP runs very well on it, and I think that 8 seconds just to read the metadata from the RAW file is not normal. So this is a big part of the time it takes to open a RAW file in AP. This operation can probably be optimized and speed up AP.

 

That's very observant. B)

 

Base on some tests (monitoring Windows processes), it appears that Photo starts a new ExifTool process each time it reads metadata from an image, then kills the process when it's done. If that's the case, there's a significant performance hit. ExifTool is written in Perl, and a new instance of ExifTool needs to start a new Perl runtime environment - this is a little expensive.

 

ExifTool has a "server" mode which allows a single ExifTool process to be started and remain resident, processing piped commands. This limits the Perl startup overhead to a one-time hit and significantly improves performance. I can personally attest to this, as I have developed my own utility software that "wraps" around ExifTool, utilizing this mode. It's much faster.

 

Maybe the Photo development team could look at implementing this? It's documented under the -stay_open FLAG section here: http://www.sno.phy.queensu.ca/~phil/exiftool/exiftool_pod.html

[Mark Ingram]

Thanks Len, that sounds interesting.

[lenmerkel]

You're welcome!