mirror of
https://github.com/python-pillow/Pillow.git
synced 2024-11-13 21:26:46 +03:00
319 lines
10 KiB
Python
319 lines
10 KiB
Python
from helper import unittest, PillowTestCase, lena, py3
|
|
|
|
import os
|
|
|
|
from PIL import Image, TiffImagePlugin
|
|
|
|
|
|
class TestFileLibTiff(PillowTestCase):
|
|
|
|
def setUp(self):
|
|
codecs = dir(Image.core)
|
|
|
|
if "libtiff_encoder" not in codecs or "libtiff_decoder" not in codecs:
|
|
self.skipTest("tiff support not available")
|
|
|
|
def _assert_noerr(self, im):
|
|
"""Helper tests that assert basic sanity about the g4 tiff reading"""
|
|
# 1 bit
|
|
self.assertEqual(im.mode, "1")
|
|
|
|
# Does the data actually load
|
|
im.load()
|
|
im.getdata()
|
|
|
|
try:
|
|
self.assertEqual(im._compression, 'group4')
|
|
except:
|
|
print("No _compression")
|
|
print (dir(im))
|
|
|
|
# can we write it back out, in a different form.
|
|
out = self.tempfile("temp.png")
|
|
im.save(out)
|
|
|
|
def test_g4_tiff(self):
|
|
"""Test the ordinary file path load path"""
|
|
|
|
file = "Tests/images/lena_g4_500.tif"
|
|
im = Image.open(file)
|
|
|
|
self.assertEqual(im.size, (500, 500))
|
|
self._assert_noerr(im)
|
|
|
|
def test_g4_large(self):
|
|
file = "Tests/images/pport_g4.tif"
|
|
im = Image.open(file)
|
|
self._assert_noerr(im)
|
|
|
|
def test_g4_tiff_file(self):
|
|
"""Testing the string load path"""
|
|
|
|
file = "Tests/images/lena_g4_500.tif"
|
|
with open(file, 'rb') as f:
|
|
im = Image.open(f)
|
|
|
|
self.assertEqual(im.size, (500, 500))
|
|
self._assert_noerr(im)
|
|
|
|
def test_g4_tiff_bytesio(self):
|
|
"""Testing the stringio loading code path"""
|
|
from io import BytesIO
|
|
file = "Tests/images/lena_g4_500.tif"
|
|
s = BytesIO()
|
|
with open(file, 'rb') as f:
|
|
s.write(f.read())
|
|
s.seek(0)
|
|
im = Image.open(s)
|
|
|
|
self.assertEqual(im.size, (500, 500))
|
|
self._assert_noerr(im)
|
|
|
|
def test_g4_eq_png(self):
|
|
""" Checking that we're actually getting the data that we expect"""
|
|
png = Image.open('Tests/images/lena_bw_500.png')
|
|
g4 = Image.open('Tests/images/lena_g4_500.tif')
|
|
|
|
self.assert_image_equal(g4, png)
|
|
|
|
# see https://github.com/python-pillow/Pillow/issues/279
|
|
def test_g4_fillorder_eq_png(self):
|
|
""" Checking that we're actually getting the data that we expect"""
|
|
png = Image.open('Tests/images/g4-fillorder-test.png')
|
|
g4 = Image.open('Tests/images/g4-fillorder-test.tif')
|
|
|
|
self.assert_image_equal(g4, png)
|
|
|
|
def test_g4_write(self):
|
|
"""Checking to see that the saved image is the same as what we wrote"""
|
|
file = "Tests/images/lena_g4_500.tif"
|
|
orig = Image.open(file)
|
|
|
|
out = self.tempfile("temp.tif")
|
|
rot = orig.transpose(Image.ROTATE_90)
|
|
self.assertEqual(rot.size, (500, 500))
|
|
rot.save(out)
|
|
|
|
reread = Image.open(out)
|
|
self.assertEqual(reread.size, (500, 500))
|
|
self._assert_noerr(reread)
|
|
self.assert_image_equal(reread, rot)
|
|
self.assertEqual(reread.info['compression'], 'group4')
|
|
|
|
self.assertEqual(reread.info['compression'], orig.info['compression'])
|
|
|
|
self.assertNotEqual(orig.tobytes(), reread.tobytes())
|
|
|
|
def test_adobe_deflate_tiff(self):
|
|
file = "Tests/images/tiff_adobe_deflate.tif"
|
|
im = Image.open(file)
|
|
|
|
self.assertEqual(im.mode, "RGB")
|
|
self.assertEqual(im.size, (278, 374))
|
|
self.assertEqual(
|
|
im.tile[0][:3], ('tiff_adobe_deflate', (0, 0, 278, 374), 0))
|
|
im.load()
|
|
|
|
def test_write_metadata(self):
|
|
""" Test metadata writing through libtiff """
|
|
img = Image.open('Tests/images/lena_g4.tif')
|
|
f = self.tempfile('temp.tiff')
|
|
|
|
img.save(f, tiffinfo=img.tag)
|
|
|
|
loaded = Image.open(f)
|
|
|
|
original = img.tag.named()
|
|
reloaded = loaded.tag.named()
|
|
|
|
# PhotometricInterpretation is set from SAVE_INFO,
|
|
# not the original image.
|
|
ignored = [
|
|
'StripByteCounts', 'RowsPerStrip',
|
|
'PageNumber', 'PhotometricInterpretation']
|
|
|
|
for tag, value in reloaded.items():
|
|
if tag not in ignored:
|
|
if tag.endswith('Resolution'):
|
|
val = original[tag]
|
|
self.assert_almost_equal(
|
|
val[0][0]/val[0][1], value[0][0]/value[0][1],
|
|
msg="%s didn't roundtrip" % tag)
|
|
else:
|
|
self.assertEqual(
|
|
original[tag], value, "%s didn't roundtrip" % tag)
|
|
|
|
for tag, value in original.items():
|
|
if tag not in ignored:
|
|
if tag.endswith('Resolution'):
|
|
val = reloaded[tag]
|
|
self.assert_almost_equal(
|
|
val[0][0]/val[0][1], value[0][0]/value[0][1],
|
|
msg="%s didn't roundtrip" % tag)
|
|
else:
|
|
self.assertEqual(
|
|
value, reloaded[tag], "%s didn't roundtrip" % tag)
|
|
|
|
def test_g3_compression(self):
|
|
i = Image.open('Tests/images/lena_g4_500.tif')
|
|
out = self.tempfile("temp.tif")
|
|
i.save(out, compression='group3')
|
|
|
|
reread = Image.open(out)
|
|
self.assertEqual(reread.info['compression'], 'group3')
|
|
self.assert_image_equal(reread, i)
|
|
|
|
def test_little_endian(self):
|
|
im = Image.open('Tests/images/16bit.deflate.tif')
|
|
self.assertEqual(im.getpixel((0, 0)), 480)
|
|
self.assertEqual(im.mode, 'I;16')
|
|
|
|
b = im.tobytes()
|
|
# Bytes are in image native order (little endian)
|
|
if py3:
|
|
self.assertEqual(b[0], ord(b'\xe0'))
|
|
self.assertEqual(b[1], ord(b'\x01'))
|
|
else:
|
|
self.assertEqual(b[0], b'\xe0')
|
|
self.assertEqual(b[1], b'\x01')
|
|
|
|
out = self.tempfile("temp.tif")
|
|
# out = "temp.le.tif"
|
|
im.save(out)
|
|
reread = Image.open(out)
|
|
|
|
self.assertEqual(reread.info['compression'], im.info['compression'])
|
|
self.assertEqual(reread.getpixel((0, 0)), 480)
|
|
# UNDONE - libtiff defaults to writing in native endian, so
|
|
# on big endian, we'll get back mode = 'I;16B' here.
|
|
|
|
def test_big_endian(self):
|
|
im = Image.open('Tests/images/16bit.MM.deflate.tif')
|
|
|
|
self.assertEqual(im.getpixel((0, 0)), 480)
|
|
self.assertEqual(im.mode, 'I;16B')
|
|
|
|
b = im.tobytes()
|
|
|
|
# Bytes are in image native order (big endian)
|
|
if py3:
|
|
self.assertEqual(b[0], ord(b'\x01'))
|
|
self.assertEqual(b[1], ord(b'\xe0'))
|
|
else:
|
|
self.assertEqual(b[0], b'\x01')
|
|
self.assertEqual(b[1], b'\xe0')
|
|
|
|
out = self.tempfile("temp.tif")
|
|
im.save(out)
|
|
reread = Image.open(out)
|
|
|
|
self.assertEqual(reread.info['compression'], im.info['compression'])
|
|
self.assertEqual(reread.getpixel((0, 0)), 480)
|
|
|
|
def test_g4_string_info(self):
|
|
"""Tests String data in info directory"""
|
|
file = "Tests/images/lena_g4_500.tif"
|
|
orig = Image.open(file)
|
|
|
|
out = self.tempfile("temp.tif")
|
|
|
|
orig.tag[269] = 'temp.tif'
|
|
orig.save(out)
|
|
|
|
reread = Image.open(out)
|
|
self.assertEqual('temp.tif', reread.tag[269])
|
|
|
|
def test_12bit_rawmode(self):
|
|
""" Are we generating the same interpretation
|
|
of the image as Imagemagick is? """
|
|
TiffImagePlugin.READ_LIBTIFF = True
|
|
# Image.DEBUG = True
|
|
im = Image.open('Tests/images/12bit.cropped.tif')
|
|
im.load()
|
|
TiffImagePlugin.READ_LIBTIFF = False
|
|
# to make the target --
|
|
# convert 12bit.cropped.tif -depth 16 tmp.tif
|
|
# convert tmp.tif -evaluate RightShift 4 12in16bit2.tif
|
|
# imagemagick will auto scale so that a 12bit FFF is 16bit FFF0,
|
|
# so we need to unshift so that the integer values are the same.
|
|
|
|
im2 = Image.open('Tests/images/12in16bit.tif')
|
|
|
|
if Image.DEBUG:
|
|
print (im.getpixel((0, 0)))
|
|
print (im.getpixel((0, 1)))
|
|
print (im.getpixel((0, 2)))
|
|
|
|
print (im2.getpixel((0, 0)))
|
|
print (im2.getpixel((0, 1)))
|
|
print (im2.getpixel((0, 2)))
|
|
|
|
self.assert_image_equal(im, im2)
|
|
|
|
def test_blur(self):
|
|
# test case from irc, how to do blur on b/w image
|
|
# and save to compressed tif.
|
|
from PIL import ImageFilter
|
|
out = self.tempfile('temp.tif')
|
|
im = Image.open('Tests/images/pport_g4.tif')
|
|
im = im.convert('L')
|
|
|
|
im = im.filter(ImageFilter.GaussianBlur(4))
|
|
im.save(out, compression='tiff_adobe_deflate')
|
|
|
|
im2 = Image.open(out)
|
|
im2.load()
|
|
|
|
self.assert_image_equal(im, im2)
|
|
|
|
def test_compressions(self):
|
|
im = lena('RGB')
|
|
out = self.tempfile('temp.tif')
|
|
|
|
for compression in ('packbits', 'tiff_lzw'):
|
|
im.save(out, compression=compression)
|
|
im2 = Image.open(out)
|
|
self.assert_image_equal(im, im2)
|
|
|
|
im.save(out, compression='jpeg')
|
|
im2 = Image.open(out)
|
|
self.assert_image_similar(im, im2, 30)
|
|
|
|
def test_cmyk_save(self):
|
|
im = lena('CMYK')
|
|
out = self.tempfile('temp.tif')
|
|
|
|
im.save(out, compression='tiff_adobe_deflate')
|
|
im2 = Image.open(out)
|
|
self.assert_image_equal(im, im2)
|
|
|
|
def xtest_bw_compression_wRGB(self):
|
|
""" This test passes, but when running all tests causes a failure due
|
|
to output on stderr from the error thrown by libtiff. We need to
|
|
capture that but not now"""
|
|
|
|
im = lena('RGB')
|
|
out = self.tempfile('temp.tif')
|
|
|
|
self.assertRaises(
|
|
IOError, lambda: im.save(out, compression='tiff_ccitt'))
|
|
self.assertRaises(IOError, lambda: im.save(out, compression='group3'))
|
|
self.assertRaises(IOError, lambda: im.save(out, compression='group4'))
|
|
|
|
def test_fp_leak(self):
|
|
im = Image.open("Tests/images/lena_g4_500.tif")
|
|
fn = im.fp.fileno()
|
|
|
|
os.fstat(fn)
|
|
im.load() # this should close it.
|
|
self.assertRaises(OSError, lambda: os.fstat(fn))
|
|
im = None # this should force even more closed.
|
|
self.assertRaises(OSError, lambda: os.fstat(fn))
|
|
self.assertRaises(OSError, lambda: os.close(fn))
|
|
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|
|
|
|
# End of file
|