First thanks again to all you good people for providing and contributing to this great piece of software. I'd like to offer some small suggestion b/c even after rtfm'ing quite a lot got tricked by the various IM versions and think more verbosity on this might be helpful to us simple users:
Some operations - for example -colorspace RGB -yadayadayada -colorspace sRGB results in severe image degradation when used with Q8, and some operations are advised to be only employed with a hdri build. If this is all known it would be totally great to add warnings to respective IM ops to notifiy the user that he/she is probably doing something he/she isn't entirely aware of. That's because if you simply look up the command options, you don't stumble about this problem even though it is scattered around other docs pages on the various im sites.
So as long as you provide Q8/Q16 builds for speed (and you might even with im7 after it's final), is adding a quick do_quantum_check_warning(16) to the code a lot of hassle? It probably doesn't concern that many operations, but I wouldn't know 'bout that. Btw the same goes for filters that are supposed to be used with -distort and not -resize.
Suggestion: "too low Quantum" warning
Re: Suggestion: "too low Quantum" warning
It probably would concern many operations. People would soon get tired of seeing the warning and ask us for a way to defeat it or to disable it by default.
Re: Suggestion: "too low Quantum" warning
Interesting, so that would make a warning even more useful than I imagined as Q8 is such a severe downgrade to iq and unknown to a lot of ppl. Currently, (at least for me, ymmv) it's very hard to get an idea when Q8 can be used for speed at all w/o any docs or warnings.glennrp wrote:It probably would concern many operations.
It's not like a -quiet command line option is unknown in the World Of Software, suppressing non-error output.glennrp wrote:People would soon get tired of seeing the warning and ask us for a way to defeat it or to disable it by default.
But of course your point is taken, as a dev I wouldn't want to hassle with work, too, that doesn't fix bugs and only affects people that aren't experts or have lots time to rtfm ... but being affected by this problem, I wanted to mention the suggestion anyway.
-
snibgo
- Posts: 12159
- Joined: 2010-01-23T23:01:33-07:00
- Authentication code: 1151
- Location: England, UK
Re: Suggestion: "too low Quantum" warning
In principle I think it's a good idea. But working out the practicalities is hard.
Almost any operation, at any Q-number, removes some data. How much loss is considered bad? The sRGB->RGB->sRGB round-trip in Q8 is bad for shadows, but if the image has no shadows there is no (or hardly any) problem.
An aspect that I often consider is: how much data loss is there? I often measure this by comparing the number of unique values input with the number output. But a good measure isn't that simple. A 50% data loss isn't a problem if I start with 2^32 values, and probably not if I have only 2^16 values. Even 50% loss of 2^8 values, is spread evenly across a "typical" web photo, may not be a problem.
This ties into a wider issue that I've always wanted: a "lint" processor within IM that would tell a user, "Your command is legal, but probably not what you want." The practicalities of this are also hard.
Personally, I use Q8 only for graphics, never for photos.
Almost any operation, at any Q-number, removes some data. How much loss is considered bad? The sRGB->RGB->sRGB round-trip in Q8 is bad for shadows, but if the image has no shadows there is no (or hardly any) problem.
An aspect that I often consider is: how much data loss is there? I often measure this by comparing the number of unique values input with the number output. But a good measure isn't that simple. A 50% data loss isn't a problem if I start with 2^32 values, and probably not if I have only 2^16 values. Even 50% loss of 2^8 values, is spread evenly across a "typical" web photo, may not be a problem.
This ties into a wider issue that I've always wanted: a "lint" processor within IM that would tell a user, "Your command is legal, but probably not what you want." The practicalities of this are also hard.
Personally, I use Q8 only for graphics, never for photos.
snibgo's IM pages: im.snibgo.com
Re: Suggestion: "too low Quantum" warning
Ok, fair enough, I see you've already considered the problemsnibgo wrote:In principle I think it's a good idea. But working out the practicalities is hard.This ties into a wider issue that I've always wanted: a "lint" processor within IM that would tell a user, "Your command is legal, but probably not what you want." The practicalities of this are also hard.
so for the basics I'd personally strongly suggest adding a bit fat warning to ...1) the download page about image degradation on photos on god-knows-how-many ops
2) the -colorspace docs page, too as linear light is suggested/used in lotsa scripts. Strangely, there already is a warning about hdri ("Note the scRGB colorspace requires HDRI support otherwise it behaves just like linear RGB), so that really cons you into believing q8/q16 doesn't matter: http://www.imagemagick.org/script/comma ... colorspace
Now I know you guys are all IM wizards and maybe it's hard to get down to the level of your basic IM newbie user, but as long as you provide q8 builds (which is useful as they *are* faster and that's why I use 'em for batch ops) this really a bad trap as you won't realize the img degradation if you don't look closely but trust IM to get it right. Same probably goes for hdri, btw.