Demo: Extracting curve data from PDFs

The notebook demonstrates how to extract curve objects from PDFs using pdfplumber.



In [1]:

    
import pdfplumber

Load the PDF

Here, we're using the USDA's "National Weekly Ag Energy Round-Up", a weekly one-page report.



In [2]:

    
report = pdfplumber.open("../pdfs/ag-energy-round-up-2017-02-24.pdf").pages[0]



In [3]:

    
im = report.to_image()
im









    Out[3]:

Examine the `curve` objects

Page objects have a .curves property, which contains a list of curve objects found on the page. This report contains 12 curves, four per chart:



In [4]:

    
len(report.curves)









    Out[4]:





12

Here's what the first curve object looks like:



In [5]:

    
report.curves[0]









    Out[5]:





{'bottom': Decimal('168.720'),
 'doctop': Decimal('59.740'),
 'height': Decimal('108.980'),
 'linewidth': Decimal('0.250'),
 'object_type': 'curve',
 'page_number': 1,
 'points': [(Decimal('583.710'), Decimal('168.720')),
  (Decimal('749.790'), Decimal('168.720')),
  (Decimal('583.710'), Decimal('155.160')),
  (Decimal('749.790'), Decimal('155.160')),
  (Decimal('583.710'), Decimal('141.480')),
  (Decimal('749.790'), Decimal('141.480')),
  (Decimal('583.710'), Decimal('127.920')),
  (Decimal('749.790'), Decimal('127.920')),
  (Decimal('583.710'), Decimal('114.240')),
  (Decimal('749.790'), Decimal('114.240')),
  (Decimal('583.710'), Decimal('100.680')),
  (Decimal('749.790'), Decimal('100.680')),
  (Decimal('583.710'), Decimal('87.000')),
  (Decimal('749.790'), Decimal('87.000')),
  (Decimal('583.710'), Decimal('73.320')),
  (Decimal('749.790'), Decimal('73.320')),
  (Decimal('583.710'), Decimal('59.740')),
  (Decimal('749.790'), Decimal('59.740'))],
 'top': Decimal('59.740'),
 'width': Decimal('166.080'),
 'x0': Decimal('583.710'),
 'x1': Decimal('749.790'),
 'y0': Decimal('443.280'),
 'y1': Decimal('552.260')}

Display the curves

We can get a quick sense of where the curves are by passing them to PageImage.draw_lines:



In [6]:

    
im.draw_lines(report.curves, stroke="red", stroke_width=2)









    Out[6]:

We can get a more better sense of the curves by cycling through a four-color palette:



In [7]:

    
im.reset()
colors = [ "gray", "red", "blue", "green" ]
for i, curve in enumerate(report.curves):
    stroke = colors[i%len(colors)]
    im.draw_circles(curve["points"], radius=3, stroke=stroke, fill="white")
    im.draw_line(curve["points"], stroke=stroke, stroke_width=2)
im









    Out[7]:

Note: Above, you'll notice the zig-zag pattern made by the curve that describes the gridlines. That's because pdfminer (and, hence, pdfplumber) currently only provide access to a the points on a curve, and not the actual path of the curve. The actual path can — as with the gridlines — include both "lineto" and "moveto" commands.

Demo: Extracting curve data from PDFs

Load the PDF

Examine the curve objects

Display the curves

Examine the `curve` objects