Even if it is not C code, it is faster than those compiled functions (at least it was the last time I checked about a year ago when I started my package).
I guess that Exif has improved pretty much since then. Perhaps Marcus
can clue us in, whether it's possible that a userland implementation is
faster than his extension?
After doing some benchmarks against exif_read_data in 4.3, I got to admit it has improved a lot :-) So I no longer can say my code is faster. In fact... it seems that exif_read_data now runs about 20 times faster.
It might be possible for Image_JPEG to wrap exif_read_data when parsing files for read-only purposes, but I wouldn't get into that at this point.
Image_JPEG can process about 100 images per second on a 1GHz Pentium.
Can you, as a first step, perhaps have a look at Image_IPTC and
summarize, how the packages differ?
I would... if Image_IPTC would have been released. I didn't noticed that when I first saw it on the packages database.
I sent an email to it's author.
However, I think having a single Image_JPEG module (instead of Image_EXIF and Image_IPTC) would be a better idea.
Regards,
Sebastian