Fix test suite when reading image file

[deeplook]

Looks like after moving the test suite one level up some images for the W3C SVG tests cannot be found anymore. Same on Py2 and 3.

svglib $ py.test -v -s tests/test_samples.py
[...]
============================================================= FAILURES =============================================================
___________________________________________________ TestW3CSVG.test_convert_pdf ____________________________________________________

self = <test_samples.TestW3CSVG object at 0x108d41a10>

    def test_convert_pdf(self):
        "Test converting W3C SVG files to PDF using svglib."

        exclude_list = [
            "paint-stroke-06-t.svg",
            "coords-trans-09-t.svg",  # renderPDF issue (div by 0)
            # ExpatError while parsing due to the ev namespace
            "interact-order-04-t.svg",
            "interact-order-05-t.svg",
            "media-video-220-t.svg",
            "script-listener-201-t.svg",
            # Unsupported 'transform="ref(svg, ...)"' expression
            "coords-constr-201-t.svg",
            "coords-constr-202-t.svg",
            "coords-constr-203-t.svg",
            "coords-constr-204-t.svg",
            # Errors to be investigated
            "udom-event-207-t.svg",
        ]

        paths = glob.glob("%s/svg/*.svg" % self.folder_path)
        msg = "Destination folder '%s/svg' not found." % self.folder_path
        assert len(paths) > 0, msg

        for i, path in enumerate(paths):
            print("working on [%d] %s" % (i, path))

            if basename(path) in exclude_list:
                print("excluded (to be tested later)")
                continue

            # convert
            drawing = svglib.svg2rlg(path)

            # save as PDF
            base = splitext(path)[0] + '-svglib.pdf'
>           renderPDF.drawToFile(drawing, base, showBoundary=0)

tests/test_samples.py:428:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/graphics/renderPDF.py:288: in drawToFile
    draw(d, c, 0, 0, showBoundary=showBoundary)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/graphics/renderPDF.py:29: in draw
    R.draw(renderScaledDrawing(drawing), canvas, x, y, showBoundary=showBoundary)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/graphics/renderbase.py:199: in draw
    self.drawNode(drawing)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/graphics/renderPDF.py:55: in drawNode
    self.drawNodeDispatcher(node)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/graphics/renderbase.py:280: in drawNodeDispatcher
    self.drawGroup(node)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/graphics/renderbase.py:309: in drawGroup
    self.drawNode(node)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/graphics/renderPDF.py:55: in drawNode
    self.drawNodeDispatcher(node)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/graphics/renderbase.py:280: in drawNodeDispatcher
    self.drawGroup(node)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/graphics/renderbase.py:309: in drawGroup
    self.drawNode(node)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/graphics/renderPDF.py:55: in drawNode
    self.drawNodeDispatcher(node)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/graphics/renderbase.py:280: in drawNodeDispatcher
    self.drawGroup(node)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/graphics/renderbase.py:309: in drawGroup
    self.drawNode(node)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/graphics/renderPDF.py:55: in drawNode
    self.drawNodeDispatcher(node)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/graphics/renderbase.py:264: in drawNodeDispatcher
    self.drawImage(node)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/graphics/renderPDF.py:87: in drawImage
    image.width, image.height
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/pdfgen/canvas.py:852: in drawInlineImage
    img_obj = PDFImage(image, x,y, width, height)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/pdfgen/pdfimages.py:40: in __init__
    self.getImageData()
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/pdfgen/pdfimages.py:156: in getImageData
    imagedata, imgwidth, imgheight = self.non_jpg_imagedata(image)  #try for normal kind of image
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/pdfgen/pdfimages.py:138: in non_jpg_imagedata
    imagedata = pdfutils.cacheImageFile(image,returnInMemory=1)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/pdfbase/pdfutils.py:95: in cacheImageFile
    code = makeA85Image(filename,IMG)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/pdfbase/pdfutils.py:38: in makeA85Image
    raw = img.getRGBData()
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/lib/utils.py:883: in getRGBData
    annotateException('\nidentity=%s'%self.identity())
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/lib/utils.py:1387: in annotateException
    rl_reraise(t,v,b)
/Users/dinu/miniconda2/lib/python2.7/site-packages/reportlab/lib/utils.py:880: in getRGBData
    self._data = (im.tobytes if hasattr(im, 'tobytes') else im.tostring)()  #make pillow and PIL both happy, for now
/Users/dinu/miniconda2/lib/python2.7/site-packages/PIL/Image.py:678: in tobytes
    self.load()
/Users/dinu/miniconda2/lib/python2.7/site-packages/PIL/ImageFile.py:235: in load
    raise_ioerror(e)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

error = -2

    def raise_ioerror(error):
        try:
            message = Image.core.getcodecstatus(error)
        except AttributeError:
            message = ERRORS.get(error)
        if not message:
            message = "decoder error %d" % error
>       raise IOError(message + " when reading image file")
E       IOError: broken data stream when reading image file
E       identity=[ImageReader@0x1099eaad0 filename=u'/Users/dinu/repos/github/deeplook/svglib/tests/samples/W3C_SVG_12_TinyTestSuite/svg/../images/cloudqllo.jpg']

/Users/dinu/miniconda2/lib/python2.7/site-packages/PIL/ImageFile.py:59: IOError
========================================== 1 failed, 3 passed, 4 skipped in 40.64 seconds ==========================================

[claudep]

I can't open that image (cloudqllo.jpg) in my image viewers (Gimp cannot open it either). So I'm rather surprised the test passed before! We might have to find a way to know if the image is valid before rendering.

[deeplook]

Same here, and this image is not even referenced explicitly in any of the SVG files in W3C_SVG_12_TinyTestSuite/svg. So maybe it's used in some kind of loop...

[claudep]

I'm working on convertImage now, several things need updating.

[claudep]

With the above commits, the tests run fine on Python 2, however on Python 3, I'm confronted with https://bitbucket.org/rptlab/reportlab/issues/98/

[deeplook]

Funny error. Especially since the reportlab.graphics.renderbase.Renderer class has neither a drawImage method nor a superclass...

[deeplook]

@claudep I suggest you send a message on the RL mailing list, which is what they observe more thoroughly than some issue tracker. And tell Robin this is kind of a late X-mas gift from me. ;-)

[claudep]

I don't know if this is really a gift or not for RL, as I don't know the capabilities of their closed-source PLUS version.

[claudep]

ReportLab 3.3.26 solves this issue. Now should we still add some workaround code with Python 3/ReportLab < 3.3.26, considering that such combinations might be used in the next two years at least? (next Debian will have 3.3.0, as has Ubuntu 16.04).

[claudep]

However, I'm not sure we can do anything, as we cannot know whether the drawing we are building will be used to render an image or a PDF or anything else.

[deeplook]

Assuming reportlab 3.3.26 can be installed flawlessly, I'd opt for adding a note to the testsuite's README and the respective code line where the PNGs are created to use this with reportlab 3.3.36, rather than pin this version in the requirements, since I believe typical reportlab users are not jumping on new versions automatically. Does that make sense?

[claudep]

+1 to not require this version and to a note in the README.
I don't think converting to PNG is a much used ability of svglib, considering ImageMagick does it too.

[deeplook]

Done in d122b5b.