Fix black line appearing in some images

[tschaef]

Restore the previous size calculation to prevent a black horizontal line appearing in some images. This was caused by the height being calculated as one pixel too short on certain zoom levels.

[mejackreed]

Thanks for the contribution @tschaef! Could you provide some examples so we can test against this in the future?

[mejackreed]

This image url request would now create a non-canonical image api request.

Could you provide some examples where you are seeing this so we could address the issue at a larger level? I'm concerned about removing canonical urls as it would not allow Level 0 consumers.

[tschaef]

Hi @mejackreed, thanks for the quick reply and the great work with the plugin. I will try to identify why the problem only appears with certain images and get back to you with some examples.

[tschaef]

Sorry for the long wait, I was travelling. Could you explain how this would create non-canonical requests? The IIIF API allows to provide either width or height or both, even for Level 0, or did I miss something here?

[mejackreed]

Sure thing, the canonical uri syntax is described here: http://iiif.io/api/image/2.1/#canonical-uri-syntax

From that document in the size section:

the w, syntax for images that should be scaled maintaining the aspect ratio,
and the w,h syntax for explicit sizes that change the aspect ratio.

As I read this PR, it seems like it introduces size requests that have ,h which would not conform to that. Since Leaflet-IIIF doesn't modify any aspect ratios, we aren't going to support the w,h parameters either.

The behavior you are describing I believe has been removed in the v2.1 image API. See: http://iiif.io/api/image/2.1/compliance/#size

cc @azaroth42 to make sure I'm reasoning correctly

[tschaef]

I understood this as just an example, since ,h is still a valid size in v2.1.

As for a demo showing the black line: Check the Leaflet-IIIF example page. With the lowest zoom setting, it appears below the image (at least in current Chrome and Firefox). When using tiles, it could appear right in the middle.

[mejackreed]

Thanks @tschaef . The way I understand the canonical uri is that w, should only be requested. This allows adopters to pregenerate tiles nicely. See http://mejackreed.github.io/Leaflet-IIIF/examples/example.html "A static tile source".

As far as the line goes, I think this is due to rounding implementation differences in Leaflet-IIIF and the server. It seems like using Math.round instead of Math.ceil may solve this problem in both cases. What do you think?

[tschaef]

Unfortunately, this does not work, since the height could still assume non-integer values with no control over how they are rounded. Maybe @azaroth42 could actually give us a hint?

[tschaef]

The issue seems to be fixed, closing this.