Screen Capturing IDL GUIs From Code
Customers have long requested the ability to take screen captures of IDL widgets programmatically from within the context of IDL itself.
Alas, it would appear we aren't going to get that feature anytime soon, so I've been scrounging the internets for possible work-arounds.
On Linux systems, it's long been possible to simply SPAWN off a call to imagemagick's "import" command line application.
You can make screen captures of the full display, or of individually-named windows if the imagemagick distribution is in your path.
The screen scrape is stored to a temporary file, which you can then read into IDL. The following assumes that imagemagick's "import" executable is in your shell's path.
IDL> spawn, 'import -window root screenshot.jpg', output, error
IDL> im = read_jpeg('screenshot.jpg')
IDL> i = image(im)
IDL> file_delete, 'screenshot.jpg'
Imagemagick has been around for a long time. Trust it as far as you can or are allowed to by your system administrator.
If you are a Linux user, this is a strong candidate for a viable solution and doesn't require modifications to the IDL internals to support it.
On Windows, imagemagick apparently will serve this purpose if you have a Cygwin X server running as well.
But who wants to deal with that when there are simpler solutions?
I recently ran across a simple DOS .bat script with embedded C# by Vasil Arnaudov available on github that gives Windows users the equivalent functionality, without the need to install "freeware" of suspect origin, intentions, or copyright restrictions.
You can read the source and decide if you trust it or not.
The only requirement is that you've installed .NET which is nearly impossible NOT to have done if you're on any Windows version newer than XP.
In this example, I've made a copy of Vasil Arnaudov's "screenCapture.bat" to my Windows Desktop before executing it.
IDL> spawn, 'c:\users\me\desktop\screenCapture.bat test.png', output, error
IDL> im = read_png('test.png')
IDL> i = image(im[0:2, *, *])
IDL> file_delete, 'test.png'
Because I chose to capture a PNG, I picked up the transparency channel as well, therefore the first dimension is 4 elements (RGBA) instead of 3 (RGB).
You might choose to capture in a different format such as JPEG or BMP. The file format options can be found by perusing the source code.
Let's say I just want to capture the contents of a single widget.
IDL> tlb = widget_base(/column, xoffset=100, yoffset=200, title='my widget')
IDL> l = widget_label(tlb, value='We will capture the contents of this widget')
IDL> widget_control, tlb, /realize
Feel free to move the widget to a new location on the screen before capturing the full screen.
IDL> spawn, 'c:\users\me\desktop\screenCapture.bat test.jpg', output, error
IDL> read_jpeg, 'test.jpg', im
IDL> i2 = image(im)
IDL> file_delete, 'test.jpg'
In theory, screenCapture.bat will accept the name of a window as an optional argument and only capture that. But in my testing with IDL, I found that focus often (but not always) returned to the main IDL Workbench and the capture was on that window instead of the widget I had intended.
How, then, can we extract only the contents of our widget from the full-screen capture?
Keeping in mind that our screen scrape image's origin is at the lower left corner but our screen offsets are measured from the upper left, use the geometry information structure returned from WIDGET_INFO to locate and extract only a bitmap of the widget.
IDL> g = widget_info(tlb, /geometry)
IDL> s = get_screen_size()
IDL> ox = g.xoffset
IDL> oy = s - g.yoffset
IDL> w = im[*, ox:ox + g.scr_xsize, (oy - g.scr_ysize):oy]
IDL> i = image(w)
If you are working with multiple monitors and wish to capture images from a monitor other than the primary monitor, you may need to modify the C# code that's provided, an exercise left for the reader.
This logic could be encapsulated in a function quite easily, of course, one that accepts a widget ID as input and returns an image array as output. This is also left as an exercise for the reader.