New stable version, 7.36.0

There's yet another new version of vips. Here's what's new in headlines, see the libvips ChangeLog and nip2 ChangeLog for more detail. Sources, plus binaries for Windows are here, OS X support should be coming via homebrew soon.

Tab indicators in nip2

nip2 has indicators in tab labels: they show if a tab has been locked and if a tab contains an error:

Here, marklung contains an error (the red alert sign) and image, the current tab, has been locked and everything is greyed out. Locked tabs are handy for large workspaces you share with other people, it's a visual indication that this tab is read-only. There's an item in the Edit menu to lock and unlock tabs.

Reduced memory use

libvips has a new IO mode called SEQUENTIAL_UNBUFFERED which gives a useful reduction in memory use in some common cases. For example, under 7.34 we had:

$ vips copy wtc.png wtc.jpg --vips-leak --vips-concurrency=1
memory: high-water mark 35.13 MB

And with 7.36 we now have:

$ vips copy wtc.png wtc.jpg --vips-leak --vips-concurrency=1
memory: high-water mark 14.10 MB

Simple command-line rotate and scale

Rotating and scaling images from the command-line used to be slightly tricky: you had to calculate an affine transform matrix. A new operation, similarity, lets you specify the transform in terms of rotate and scale. For example:

vips similarity sarto_ng.v x.v --angle 20 --scale 1.1

will rotate by 20 degrees anticlockwise and scale by 10%. 

Improved file format support

libvips can now read and write webp, the new-ish graphics format for the web from Google.

The TIFF reader and writer has been heavily revised and now supports many more versions of the "standard". Previously the reader would only read image typess I had samples of and refused anything else. Now it tries to be as flexible as possible, but with a series of safety checks to hopefully prevent exploitation. It'll even read images split into separate planes, finally. The downside of this is that now you can write things like 20-channel double-precision complex float tiled TIFFs, which I'm fairly confident no other program on earth will be able to read.

The DeepZoom writer now responds to the .dz suffix. This means you can create pyramids as the output of other operations, for example:

vips shrink huge.tif pyramid.dz[layout=google] 16 16

will shrink the huge image by a factor of 16, and then create a pyramid from that using the Google tile naming convention. In effect, this would remove the bottom four layers of the pyramid.

The jpeg writer now supports the optimize_coding option. This will typically make jpeg images which are 5% smaller, at the cost of slightly slower write. For example:

$ vips copy wtc.png wtc.jpg
$ vips copy wtc.png wtc2.jpg[optimize_coding]
$ ls -l wtc*.jpg
-rw-rw-r-- 1 john john 14814016 Oct  3 13:44 wtc2.jpg
-rw-r--r-- 1 john john 15132795 Oct  3 13:43 wtc.jpg

Thank you to Lovell Fuller for this feature.

libvips rewrite continues

We are slowly continuing the complete rewrite for vips8. This time it was the histogram functions: they've all been reworked and given a nice new interface. Convolution and morphology next, probably.