image degredation over repeated steps (invert) about image-sequencer HOT 16 CLOSED

That's a really awesome idea!
But in this case, the problem is a bit different.

I just tried this:

var r = savePixels(pixels, options.format, {quality: 100});
...

That is, explicitly asking save-pixels to save in high quality.
After that, there was no deterioration as we can see above at all.

But that created a new problem. The build failed.
One unit test failed.

As you can see in the screenshot, the two values which tape says are "not equal" are actually equal.

I even tried equating them in JS. Still unequal. I assure you, they are equal. Because of this, I had to remove the {quality: 100} parameter. And hence we are facing that issue.

from image-sequencer.

Why do you think the reason could be?
For two strings to be unequal but seem equal when printed?
Different encoding?
Or What could it possibly be? I am just out of thoughts :/

from image-sequencer.

from image-sequencer.

Sorry, this somehow skipped me. Thanks for your suggestion about the String() but that didnt work. Based on StackOverflow answers, I have tried string.trim(), that didn't work either.

Also, this works:
string1.substring(0,string1.length-25) == string2.substring(0,string2.length-25) //true

If 25 is replaced with any number less than 25, the above returns false.

These are the two strings:
(When I compare them on http://www.textdiff.com/ It says that there's nothing common in the two strings! It also shows similar results -- the '25' thing)

data:image/jpeg;base64,/9j/4AAQSkZJRgABAQAAAQABAAD/2wCEAAEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAQEBAf/AABEIABAAEAMBEQACEQEDEQH/xAGiAAABBQEBAQEBAQAAAAAAAAAAAQIDBAUGBwgJCgsQAAIBAwMCBAMFBQQEAAABfQECAwAEEQUSITFBBhNRYQcicRQygZGhCCNCscEVUtHwJDNicoIJChYXGBkaJSYnKCkqNDU2Nzg5OkNERUZHSElKU1RVVldYWVpjZGVmZ2hpanN0dXZ3eHl6g4SFhoeIiYqSk5SVlpeYmZqio6Slpqeoqaqys7S1tre4ubrCw8TFxsfIycrS09TV1tfY2drh4uPk5ebn6Onq8fLz9PX29/j5+gEAAwEBAQEBAQEBAQAAAAAAAAECAwQFBgcICQoLEQACAQIEBAMEBwUEBAABAncAAQIDEQQFITEGEkFRB2FxEyIygQgUQpGhscEJIzNS8BVictEKFiQ04SXxFxgZGiYnKCkqNTY3ODk6Q0RFRkdISUpTVFVWV1hZWmNkZWZnaGlqc3R1dnd4eXqCg4SFhoeIiYqSk5SVlpeYmZqio6Slpqeoqaqys7S1tre4ubrCw8TFxsfIycrS09TV1tfY2dri4+Tl5ufo6ery8/T19vf4+fr/2gAMAwEAAhEDEQA/APxnr/Kc/wC/gKACgAoA/wD/2Q==

data:image/jpeg;base64,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

from image-sequencer.

Aha! I looked really carefully -- the first string ends with:

APxnr/Kc/wC/gKACgAoA/wD/2Q==

The second ends with:

APxfr/Kc/wC/gKACgAoA/wD/2Q==

Notice: APxnr vs. APxfr -- !!

just one character off.

https://gist.github.com/jywarren/fdb5294effe01b908245700a38adc057/revisions

from image-sequencer.

If you re-encode again, does it stay APxfr?

from image-sequencer.

Wow those are exceptionally great observation skills there, I must say!

from image-sequencer.

No, re-encoding doesn't change anything, I tried.
I think the method I am using for testing this isn't ideal.
I doubly-invert a DataURL and equate two DataURL strings.

I should probably use a different sort of comparison. Like one with a threshold.
I don't know why doubly inverting a DataURL would give a slightly different result when generated in High Quality. Do you know any articles which would help me identify which part of a DataURL/Image Binary contain what information

from image-sequencer.

from image-sequencer.

Thanks for the update! I will do that.

from image-sequencer.

The issue that we are facing, of the unequal DataURLs, could it be the result of the lossy JPEG compression?

from image-sequencer.

Turns out I am right! I converted that image to PNG and now it works! What should be done now? I still think that comparing with a threshold should should be implemented.

from image-sequencer.

from image-sequencer.

The module you mentioned npm:image-diff is pretty heavy. I am not sure we should use it.
Maybe I can write a lightweight library for that myself.

from image-sequencer.

There were a few others, no? And we don't need to include it in the library itself; it could just be included in the test suite and under devDependencies.

https://github.com/publiclab/image-sequencer/blob/master/CONTRIBUTING.md#module-candidates -- had several diff modules. I'd like to open this issue and take slow steps in this direction; not urgent!

from image-sequencer.

oops, ok -- #83!

from image-sequencer.