having trouble using set

[kayarre]

using pyvips 2.1.8

I am doing the simple operation from a loaded .svs file

test.set("xres", 0.2525)

but when I type:

test.get("xres")

it returns 1.0

is there something I am missing?

I am planning to write the output to a simple tiff format but I am having a hard time figuring out how to include the meta data with the file.

pyvips is a great tool thank you!

After looking through the documentation there appears that the image metadata is immutable, so what is the point of the set method if can't do what a set method normally does?

say I want to add a bunch of different items as metadata?

would I do something like this

new_image = image.copy().set('icc-profile-data', new_profile).set('xres', 0.2525)

or could I do a for loop, probably need check to see if that object exists

for key, meta in meta_dict.items():
    new = new.copy().set(key, meta)

[jcupitt]

Hello @kayarre,

It's a bit confusing -- libvips has two different types of image metadata. Fields like width and xres are built in, and are properties of the underlying GObject. You can read them like this:

$ python3
Python 3.7.5rc1 (default, Oct  8 2019, 16:47:45) 
[GCC 9.2.1 20191008] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import pyvips
>>> x = pyvips.Image.new_from_file("k2.jpg")
>>> x.width
1450
>>> x.xres
0.9999999999068677

There are 10 or so of these core properties. They are immutable and you can only set them with .copy() like this:

>>> y = x.copy(xres=12)
>>> y.xres
12.0
>>> x.xres
0.9999999999068677

There's a full list here:

https://libvips.github.io/pyvips/vimage.html#pyvips.Image.copy

All other image metadata is implemented as a set of name-value pairs in a hash table attached to each image. These are very flexible: you can add any type of data. You can see everything attached to an image with .get_fields():

>>> x.get_fields()
['width', 'height', 'bands', 'format', 'coding', 'interpretation', 'xoffset', 
'yoffset', 'xres', 'yres', 'filename', 'vips-loader', 'jpeg-multiscan', 
'jpeg-chroma-subsample', 'exif-data', 'resolution-unit', 'exif-ifd0-Orientation', 
'exif-ifd0-XResolution', 'exif-ifd0-YResolution', 'exif-ifd0-ResolutionUnit', 
'exif-ifd0-YCbCrPositioning', 'exif-ifd2-ExifVersion', 
'exif-ifd2-ComponentsConfiguration', 
'exif-ifd2-FlashPixVersion', 'exif-ifd2-ColorSpace', 'exif-ifd2-PixelXDimension', 
'exif-ifd2-PixelYDimension', 'orientation']

The 10 built-in fields are listed, but after that there are a lot of others. These extra fields can be changed with get and set.

>>> x.get('orientation')
1
>>> x.set('orientation', 99)
>>> x.get('orientation')
99

But be careful changing them --- I would copy the image first, or you may have problems with caching.

You can make new fields too:

>>> x.set_type(pyvips.GValue.guint64_type, 'banana', 12)
>>> x.get('banana')
12

Makes a new 64-bit int field.

[kayarre]

Thank you @jcupitt
I was actually in the midst of adding an update. Thank you it is much clearer now.

I figured out essentially what you described. The core parts are immutable but any of the others there is no issue.

I can do this:

test = invert.copy(xres = 1000.0/(float(im.get("openslide.mpp-x"))),
                                 yres = 1000.0/(float(im.get("openslide.mpp-y"))),
                       )

and it gives the correct thing.

one weird thing though is that when I add this metadata
test.set_type(pyvips.GValue.gstr_type, "resolution-unit", "mm")

and then save to tiff file i.e.

test.tiffsave("test2.tif", compression = "VIPS_FOREIGN_TIFF_COMPRESSION_DEFLATE",
properties=True, strip=False)

vipsheader gives me this

$ vipsheader -a test2.tif 
test2.tif: 16384x16384 uchar, 1 band, b-w, tiffload
width: 16384
height: 16384
bands: 1
format: uchar
coding: none
interpretation: b-w
xoffset: 0
yoffset: 0
xres: 3960.43
yres: 3960.43
filename: test2.tif
vips-loader: tiffload
n-pages: 1
image-description: <?xml version="1.0"?>
<image xmlns="http://www.vips.ecs.soton.ac.uk//dzsave" date="2019-11-10T03:25:56.958661Z" version="8.8.1">
  <properties>
    <property>
      <name>width</name>
      <value type="gint">16384</value>
    </property>
    <property...
resolution-unit: cm
orientation: 1

Why does it show resolution-unit: cm ?

the answer is that tiffsave has a flag resunit and can be either "cm" or "in"

is there reason "mm" couldn't be added to the enum?

Another Question is that it appears that none of the meta data gets written out?

[jcupitt]

Oh no! I think you've found a bug. I pushed a fix, and I now see:

$ python3
Python 3.7.5rc1 (default, Oct  8 2019, 16:47:45) 
[GCC 9.2.1 20191008] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import pyvips
>>> x = pyvips.Image.new_from_file("k2.tif")
>>> x.get("resolution-unit")
'cm'
>>> x.set("resolution-unit", "in")
>>> x.write_to_file("x.tif")
>>> 
$ tiffinfo x.tif
TIFF Directory at offset 0x87f008 (8908808)
  Image Width: 1450 Image Length: 2048
  Resolution: 71.9988, 71.9988 pixels/inch
  Bits/Sample: 8
  Sample Format: unsigned integer
  Compression Scheme: None
  Photometric Interpretation: RGB color
  Orientation: row 0 top, col 0 lhs
  Samples/Pixel: 3
  Rows/Strip: 128
  Planar Configuration: single image plane

Thanks for pointing out this dumbness.

[jcupitt]

As a workaround, you can set the res unit from the args to tiffsave:

x.write_to_file("x.tif", resunit="inch")

[tamkaho]

Oh no! I think you've found a bug. I pushed a fix, and I now see:

$ python3
Python 3.7.5rc1 (default, Oct  8 2019, 16:47:45) 
[GCC 9.2.1 20191008] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import pyvips
>>> x = pyvips.Image.new_from_file("k2.tif")
>>> x.get("resolution-unit")
'cm'
>>> x.set("resolution-unit", "in")
>>> x.write_to_file("x.tif")
>>> 
$ tiffinfo x.tif
TIFF Directory at offset 0x87f008 (8908808)
  Image Width: 1450 Image Length: 2048
  Resolution: 71.9988, 71.9988 pixels/inch
  Bits/Sample: 8
  Sample Format: unsigned integer
  Compression Scheme: None
  Photometric Interpretation: RGB color
  Orientation: row 0 top, col 0 lhs
  Samples/Pixel: 3
  Rows/Strip: 128
  Planar Configuration: single image plane

Thanks for pointing out this dumbness.

In which version has set_type been fixed?

[jcupitt]

It's in libvips 8.9.0 onwards.