On linux, the bounding box is absolute, i.e., bounding box (5, 5, 10, 10), it means: point (5,5) and point (10,10). However, on Mac, the bounding box is additive, i.e., it means point (5,5) and point (15, 15). Ugly fix could be:

    def grab_to_file(self, filename, bbox=None, dpi=72):
        # FIXME: Should query dpi from somewhere, e.g for retina displays
        region = self.CG.CGRectMake(bbox[1], bbox[2], bbox[3]-bbox[1], bbox[4]-bbox[2]) if bbox else self.CG.CGRectInfinite

        # Create screenshot as CGImage
        image = self.CG.CGWindowListCreateImage(region,
                                                self.CG.kCGWindowListOptionOnScreenOnly, self.CG.kCGNullWindowID,
self.CG.kCGWindowImageDefault)

Installed Pillow-2.7 and pyscreenshot-0.3.3 on Python 2.7.8 32-bit on Windows 7 64-bit. pyscreenshot.grab() continued to generate error:

import pyscreenshot as pss pss.grab() Traceback (most recent call last): File "", line 1, in File "C:\Python27\ArcGIS10.3\lib\site-packages\pyscreenshot-0.3.3-py2.7.egg\pyscreenshot__init__.py", line 71, in grab return grab(to_file=False, childprocess=childprocess, backend=backend, bbox=bbox) File "C:\Python27\ArcGIS10.3\lib\site-packages\pyscreenshot-0.3.3-py2.7.egg\pyscreenshot__init_.py", line 51, in _grab backend_obj = _get_loader().selected() File "C:\Python27\ArcGIS10.3\lib\site-packages\pyscreenshot-0.3.3-py2.7.egg\pyscreenshot\loader.py", line 75, in selected self.raise_exc() File "C:\Python27\ArcGIS10.3\lib\site-packages\pyscreenshot-0.3.3-py2.7.egg\pyscreenshot\loader.py", line 83, in raise_exc raise FailedBackendError(message) pyscreenshot.loader.FailedBackendError: Install at least one backend!

Updating line 14 of pil.py to: from PIL import ImageGrab

Seemed to resolve the error. That said, not sure if that will cause issues in other places.

While easy_install it, I get this:

File "/tmp/easy_install-32xFOz/pyscreenshot-0.1.0/pyscreenshot/init.py", line 1, in # ImportError: No module named path

The first screenshot taken with PyScreenshot is fine: however, after I have taken one screenshot, when I take another, the application crashes. In PyCharm, it finishes with exit code 139. In the terminal, the output is 'Segmentation fault (core dumped)' and the application stops executing.

Allows Python3 use - see #17. Note this requires all imports to be absolute -- see https://wiki.python.org/moin/PortingPythonToPy3k#relative-imports.

…st (~70ms for 4K screenshot)

Fixes / avoids https://github.com/ponty/pyscreenshot/issues/43

N.B. you must use childprocess=False to get good performance!! Unfortunately default Python inter-thread communication/forking/etc. is painfully slow. Compare:

In [5]: %time pyscreenshot.grab(backend='pygdk3')
CPU times: user 6.38 ms, sys: 0 ns, total: 6.38 ms
Wall time: 397 ms
Out[5]: <PIL.PngImagePlugin.PngImageFile image mode=RGB size=3840x2160 at 0x7F9721DA4908>

In [7]: %time pyscreenshot.grab(backend='pygdk3', childprocess=False)
CPU times: user 59.1 ms, sys: 27.9 ms, total: 87 ms
Wall time: 102 ms
Out[7]: <PIL.Image.Image image mode=RGB size=3840x2160 at 0x7F970E57FF28>

In [8]: %time pyscreenshot.grab(backend='pygdk3', childprocess=False)
CPU times: user 36.7 ms, sys: 12.1 ms, total: 48.8 ms
Wall time: 67.9 ms
Out[8]: <PIL.Image.Image image mode=RGB size=3840x2160 at 0x7F970E5841D0>

In [9]: %time pyscreenshot.grab(backend='pygdk3', childprocess=False)
CPU times: user 26.9 ms, sys: 19.4 ms, total: 46.2 ms
Wall time: 65.3 ms
Out[9]: <PIL.Image.Image image mode=RGB size=3840x2160 at 0x7F970E584550>

Backend speedtest (which does not use childprocess=False) results:

n=10                                                                                             
------------------------------------------------------                     
pygdk3                  3.8  sec        (  383 ms per call)
pyqt5                   7.4  sec        (  741 ms per call)
imagemagick             11   sec        ( 1062 ms per call)
gnome-screenshot        8.8  sec        (  883 ms per call)

but with childprocess=False it is an order of magnitude faster than other backends ^_^. Sorry I don't have scrot/pyside/pyqt/wx/pil backends installed to compare.

Is there any reason for the backend code to be structured around writing to files? And also on a separate process? For backends that don't need it (gtk, qt, ...?), it seems performance would be hundreds/thousands of times better if operations are done in memory.

I am running Windows 10 and Python 3.4.3.

Whenever I run your full screen example:

$ python -m pyscreenshot.examples.showgrabfullscreen

I only get a partial top-left of my screen:

If I force the size of my screen with grab(bbox=(0,0,3200,1800)) I get the same image and pure black filling out the rest of the area:

Are you sure that the implementation of bbox is correct in this case,

(bbox[2] - bbox[0], bbox[3] - bbox[1], bbox[0], bbox[1])

Here you are calculating the width and height from points but the other examples are with X,Y, Width, Height

So using this you will have different behaviors from the different plugins!

Does it make sense?

I suggest to change the plugin as

(bbox[2] , bbox[3], bbox[0], bbox[1])

Thanks

Here is the situation, I got more than one screen connected to my computer. And I can not capture the second screen with pyscreenshot. Tried to use coordinates outside the first screen, but only get blackout where I thought there will be the second screen's picture.

System: Pop!_OS v19.10 python3.7

Using pyscreenshot.grab() causes an OSError using scrot v1.1.1 on my system as it tries to open a file called /tmp/pyscreenshot_scrot_alnh_dsl.png but in the system its called /tmp/pyscreenshot_scrot_alnh_dsl_000.png. This does not occur when using grab(backend=gnome-screenshot) and it does not crash using scrot 0.8 compiled from source.

I have worked around this issue by adding filename = "_000.".join(filename.split(".")) on line 24 in pyscreenshot/plugins/scrot.py but its not really pretty and one of the created files are not removed from /tmp. It looks like some naming is going very wrong somewhere?

Full traceback:

  File "/home/elias-eriksson/Dev/Project/main2.py", line 7, in <module>
    a = grab(backend="scrot")
  File "/home/elias-eriksson/Dev/Project/venv/lib/python3.7/site-packages/pyscreenshot/__init__.py", line 67, in grab
    to_file=False, childprocess=childprocess, backend=backend, bbox=bbox)
  File "/home/elias-eriksson/Dev/Project/venv/lib/python3.7/site-packages/pyscreenshot/__init__.py", line 46, in _grab
    _grab_simple, imcodec.codec, to_file, backend, bbox, filename)
  File "/home/elias-eriksson/Dev/Project/venv/lib/python3.7/site-packages/pyscreenshot/procutil.py", line 37, in run_in_childprocess
    raise e
OSError: cannot identify image file '/tmp/pyscreenshot_scrot_alnh_dsl.png'

We can not use pyscreenshot on GNOME 42, I believe because the lack of the gnome-screenshot app. Do you know some command line we can use instead? If so maybe I can PR some solution. Thanks!

There are small typos in:

• README.md
• pyscreenshot/examples/virtdisp.py
• pyscreenshot/plugins/mac_screencapture.py
• tests/fillscreen.py

Fixes:

• Should read displaying rather than diplaying.
• Should read slightly rather than slighly.
• Should read screenshots rather than screneshots.

Semi-automated pull request generated by https://github.com/timgates42/meticulous/blob/master/docs/NOTE.md

Hello,

There is some problem after type: import pyscreenshot as ImageGrab

error message: No module named 'PIL'

So i have to do: pip install Pillow

After i looked it is: init.py -> 4 line: from pyscreenshot.childproc import childprocess_backend_version childproc.py -> 6 line: from pyscreenshot.imcodec import codec imcodec.py -> 3 line: from PIL import Image

So I think it needs to be added in: setup.py, install_requires

See https://developer.apple.com/documentation/avfoundation/cameras_and_media_capture/requesting_authorization_for_media_capture_on_macos

A Mac-specific code, maybe in PIL or the Python Launcher itself, needs to be added to allow the app to request the screen capture permission BEFORE the code will work on Apple MacOS Catalina or later. As none of the other solutions work on all platforms (Linux, Win10, MacOS), it would be nice if you could get this fix pushed through (or integrated into your code if possible).

I want to convert clipboard image content into an image

I tried

from PIL import ImageGrab
im = ImageGrab.grabclipboard() is macOS and Windows only
im.save(filename)

it says ImageGrab.grabclipboard() is macOS and Windows only

I think using this library its can be possible