Robinson Analytics Ltd.
Home Python Double-Byte PDFgen Doubletalk

PDFgen API Overview

This describes what I am trying to do with the low-level drawing API exposed by the class pdfgen.Canvas. You might want to read it in tandem with the test script output.

Design Goals

The first usable API I had was the PIDDLE one. However, that led to inefficiencies in both Python code and the PDF generated. Each shape-drawing method would EITHER specify line width, color, fonts etc. directly, or use the current default ones. This means either doing a lot of work in Python to see what needs to be changed, or just always setting the graphics properties for every shape, which made a lot of PDF. We also had two layers of translation going on - from PIDDLE fonts to Postscript font names to internal names - leading to messy code. To solve this, I have produced a low-level (and this high-end) API which is completely independent of PIDDLE, the implemented piddlePDF on top of it. This API may be of interest in its own right, but is also a possible model for other formats/platforms which support paths, clipping and transforms; PostScript, Java2D and Windows spring to mind. This interface is much cleaner and aims to follow a Postscript imaging model with the following key features: Robert Kern and I did a lot of work on efficient geometry,and I am particularly happy with the Path model - Python has to do almost no work (so goes faster), and it is hard for the user to write illegal sequences of PDF.

Status and Things to Do

This is the first release of this API so there are bound to be bugs. I welcome all contributions. If you find a non-obvious bug, try to send a script which reproduces it.

I intend to add the following features shortly:

  1. Page compression options
  2. An Outline Tree
  3. Better tracking of the text cursor and graphics state
  4. Improved cross-referencing of objects in the file
I'd particularly appreciate help on the test script - adding in more tests and words so that we end up with a user guide which tries each feature and talk through it. I'll entertain feature requests, but my time is limited until January.

API overview

We'll run through the main functions by category. The categories are:
  1. Document Creation and Management
  2. Text Manipulation
  3. Line Drawing Methods
  4. Shape drawing Methods
  5. Graphics State Methods
  6. Path Drawing
  7. Clipping with Paths
  8. Coordinate Transformations
  9. Special PDF Effects

Document Creation and Management

    def __init__(self,filename,pagesize=(595.27,841.89), bottomup = 1):
		"""Creates a new document. Filename is mandatory.  You can also 
		set the page size (which defaults to A4), and whether you
		want a top-down coordinate system instead of bottom-up."""
		
    def setAuthor(self, author):
        self._doc.setAuthor(author)
        
    def setTitle(self, title):
    	"Embeds info in the PDF file"
        
    def setSubject(self, subject):
		"Embeds info in the PDF file"
    
    def showPage(self):
        """Ends the page and starts a new one.  The graphics state is reset"""

    def getPageNumber(self):
		"""Useful for numbering your documents as you go!"""
        return self._pageNumber
        
    def save(self):
        """Saves the file.  If holding data, do
        a showPage() to save them having to."""

    def setPageSize(self, size):
        """accepts a 2-tuple in points for paper size"""

    def saveState(self):
		"""Saves the current grahics state so you can undo changes later"""

    def restoreState(self):
		"""Restore to state of last saveState()"""

Text Manipulation

PDF excels at placing text. We're interested in writing it out efficiently, and keeping track of the 'text cursor' so we know when we are at the bottom of the page. Note that most of these methods correspond directly to PDF operators and involve little or no processing. The metaphor is of a current text cursor which moves across and down, and you can query the canvas to get the current position. This is defined relative to 'user space' - so if distort user space, keeping track of when you have reached the bottom of the page is your problem (for the time being). Text cursor tracking is definitely breakable now but is a priority. There are three text drawing routines: textOut(), textLine(), and textLines(). textOut() does a stringWidth() to update the cursor, so for a single isolated string on the page it is more efficient to call textLine(), which only has to move the y location down. Here's the full set of text-related methods:
    def getAvailableFonts(self):
        """Returns the list of PostScript font names available.
        Standard set now, but may grow in future with font embedding."""

    def setFont(self, psfontname, size, leading = None):
        """Sets the font.  If leading not specified, defaults to 1.2 x
        font size. Raises a readable exception if an illegal font
        is supplied.  Font names are case-sensitive!"""
    
	def getCursor(self):
        """Returns current text position in user space"""        
    
	def setTextOrigin(self, x, y):
        """Sets the text location, relative to the current graphics
        coordinate system"""

    def setTextTransform(self, a, b, c, d, e, f):
        "Like setTextOrigin, but does rotation, scaling etc."

    def stringWidth(self, text):
        "gets width of a string in the current font"
    
    def textOut(self, text):
        "prints string at current point, text cursor moves across"

    def textLine(self, text=''):
        """prints string at current point, text cursor moves down.
        Can work with no argument to simply move the cursor down."""

    def textLines(self, stuff, trim=1):
        """prints multi-line or newlined strings, moving down.  One
        comon use is to quote a multi-line block in your Python code;
        since this may be indented, by default it trims whitespace
        off each line of; set trim=0 to preserve whitespace."""

    def setCharSpace(self, charSpace):
         """Adjusts inter-character spacing"""

    def setWordSpace(self, wordSpace):
        """Adjust inter-word spacing.  This can be used
        to flush-justify text - you get the width of the
        words, and add some space between them."""

    def setHorizScale(self, horizScale):
        "Stretches text out horizontally"

    def setLeading(self, leading):
        "How far to move down at the end of a line."

    def setTextRenderMode(self, mode):
        """Set the text rendering mode.

        0 = Fill text
        1 = Stroke text
        2 = Fill then stroke
        3 = Invisible
        4 = Fill text and add to clipping path
        5 = Stroke text and add to clipping path
        6 = Fill then stroke and add to clipping path
        7 = Add to clipping path"""
        
    def setRise(self, rise):
        "Move text baseline up or down to allow superscrip/subscripts"

Line drawing methods

The following methods draw lines of various types, using the current line settings:
    def line(self, x1,y1, x2,y2):
        "As it says"       

    def lines(self, linelist):
        """As line(), but slightly more efficient for lots of them -
        one stroke operation and one less function call"""
    
    def grid(self, xlist, ylist):
        """Lays out a grid in current line style.  Suuply list of
        x an y positions."""
    
    def bezier(self, x1, y1, x2, y2, x3, y3, x4, y4):
	    "Bezier curve with the four given control points"

    def arc(self, x1,y1, x2,y2, startAng=0, extent=90):
        """Draw a partial ellipse inscribed within the rectangle x1,y1,x2,y2,
        starting at startAng degrees and covering extent degrees.   Angles
        start with 0 to the right (+x) and increase counter-clockwise.
        These should have x1<x2 and y1<y2."""
Note that bezier() is more efficient than arc(), as bezier curves are PDF primitives; PDFgen approximates an arc with up to four bezier curves, one for each quadrant.

Shape drawing methods

The following methods draw regular solid shapes. In general, you specify the geometry of the shape, and whether you want it stroked (i.e. the outline drawn in the current line style), or filled in the current fill color, or both.
    def rect(self, x, y, width, height, stroke=1, fill=0):
    def ellipse(self, x1, y1, x2, y2, stroke=1, fill=0):
    def circle(self, x_cen, y_cen, r, stroke=1, fill=0):

    def wedge(self, x1,y1, x2,y2, startAng, extent, stroke=1, fill=0):
        """Like arc, but connects to the centre of the ellipse.
        Most useful for pie charts and PacMan!"""

We should probably add roundRect(), with a radius for the corners. Other shapes may be made up in two ways: (1) using the above, together with coordinate transformations, or (2) using paths. Both are covered below.

Path drawing

Postscript, PDF, and other high-end graphics APIs have the concept of a 'path'. This is a sequence of lines and solid shapes which may be built up in steps. PDF has a separate 'path object' you can use to create these. The canvas has one method to give you a fresh path, and two to do something with one:

class Canvas...
    def beginPath(self):
        """Returns a fresh path object"""
    
    def drawPath(self, aPath, stroke=1, fill=0):
        "Draw in the mode indicated"
    
	def clipPath(self, aPath, stroke=1, fill=0):
        "clip as well as drawing - subsequent drawing
		ops wil only appear if within the area of this
		path"
    
As with the basic shapes, anything you build can be filled, stroked or both. Once you have a path, you can bild it up. It does not have to be a single closed shape; a series of squares making up a chequerboard is also a valid path. Here's an example of how they are used:
		#draw a triangle
        p = canvas.beginPath()
        p.moveTo(x1, y1)
        p.lineTo(x2, y2)
		p.lineTo(x3, y3)
		p.close()  #back to first point
        self.drawPath(p,stroke=1, fill=1) #draw and fill

Here are the things you can do with paths. Basically, you can use moveTo, lineTo and curveTo to make your own shapes, or add ready made shapes.
class PDFPathObject:
    
	def moveTo(self, x, y):
	
    def lineTo(self, x, y):
	
    def curveTo(self, x1, y1, x2, y2, x3, y3):
    
    def arc(self, x1,y1, x2,y2, startAng=0, extent=90):

    def arcTo(self, x1,y1, x2,y2, startAng=0, extent=90):
        """Like arc, but draws a line from the current point to
        the start if the start is not the current point."""

    def rect(self, x, y, width, height):

    def ellipse(self, x, y, width, height):
       
    def circle(self, x_cen, y_cen, r):
        
    def close(self):
        "draws a line back to where it started"
Paths let you do almost any shape you want. But they also offer significant performance and reuse benefits. Internally, they accumulate a bunch of Postscript page-marking operators to draw your shape, and these are only added to the page stream when you call drawPath(). So you can make a path once, and then draw it many times. This saves some Python execution time, and a lot of typing time. If you compress pages, there's a chance that the compressor will pick up on the repetition and save you space too.

Clipping with Paths

The other very neat thing you can do with paths is clipping. Here we've tuirned the word 'Python' into a clipping path, and used it as a 'mask' so that anything drawn thereafter only appears if within the path. Another common use of clipping is in chart drawing - you can limit drawing to the rectangular region you are interested in, then happily draw an entire data series even if it goes out of the chart bounds.

Image Manipulation

pdfgen uses the Python Imaging Library to read and output a very wide variety of image formats. You don't need to install PIL, but you can't use images without it. Currently we only have one image drawing function:
    def drawInlineImage(self, image, x,y, width=None,height=None):
        """Draw a PIL Image into the specified rectangle.  If width and
        height are omitted, they are calculated from the image size.
        Also allow file names as well as images.  This allows a
        caching mechanism"""
 
JPEG images are inlined directly in the file - PDF supports JPEG directly. Other formats are decoded to bitmap data by the Python Imaging Library and encoded in a special way. This can take some time (a couple of seconds for a scanned holiday photo, quite a pain for you whole photo album), so you can pass in a filename and the results will be cached in an intermediate file; subsequent builds of image-rich documents go as fast as the data can be read and written to disk. We're at work on some optional C extensions to make first-time image processing faster too. One key thing not yet done is to be able to include images ONCE at the beginning of the file, and reference them later. We'll probably have some special drawing routines which can only be called until Page 1 is written to disk; you will give an image a name, and use that to refer later. There is no challenge in adding this, just needs a spare evening :-)

Graphics State Parameters

These are all evident from a look at the test document. There are a few more that can be added as the need arises, such as other color models of fill patterns, but line drawing is fairly well sorted.
    def setLineWidth(self, width):
    def setLineCap(self, mode):
    def setLineJoin(self, mode):
    def setMiterLimit(self, limit):
    def setDash(self, array=[], phase=0):
    def setFillColorRGB(self, r, g, b):
    def setStrokeColorRGB(self, r, g, b):

Coordinate Transformation

The default coordinate system is in points (72 per inch), with the origin at the bottom left corner of the page and y increasing upwards. You can alter this with the following methods, demonstrated in testpdfgen.py:
    def transform(self, a,b,c,d,e,f):
		"For mathematicians only; the next few call this"
    def translate(self, dx, dy):
		"Move axes accordingly"
    def scale(self, x, y):
		"stretch/squeeze horizontal and/or vertical axes"
    def rotate(self, theta):
        """theta is in degrees."""
    def skew(self, alpha, beta):
		"""Skew x and y axes by angle alpha and beta (degrees)"""
While we are at it, we will reiterate two very important methods, which are used in tandem with the coordinate transformations:
    def saveState(self):
    def restoreState(self):
Right now, the only way to undo a coordinate change is saveing first then restoring after you draw. These save/restore ALL graphics state parameters. They cannot be used across multple page breaks.

Special PDF Effects

At the moment, the only real PDF-specific feature is the page transition. This was added to let me give nice presentations. The comment block explains the options available. The next priority is to add Document Outline entries.
    def addLiteral(self, s, escaped=1):
		"Directly writes PDF operators to output.  For experts only"

    def setPageTransition(self, effectname=None, duration=1, 
                        direction=0,dimension='H',motion='I'):
        """PDF allows page transition effects for use when giving
        presentations.  There are six possible effects.  You can
        just guive the effect name, or supply more advanced options
        to refine the way it works.  There are three types of extra
        argument permitted, and here are the allowed values:
            direction_arg = [0,90,180,270]
            dimension_arg = ['H', 'V']
            motion_arg = ['I','O'] (start at inside or outside)
            
        This table says which ones take which arguments:

        PageTransitionEffects = {
            'Split': [direction_arg, motion_arg],
            'Blinds': [dimension_arg],
            'Box': [motion_arg],
            'Wipe' : [direction_arg],
            'Dissolve' : [],
            'Glitter':[direction_arg]
            }
        Have fun!"""


Robinson Analytics Ltd., Wimbledon, London. email mailto:andy@robanal.demon.co.uk