1 Examples
2 PDF files
pdf-document?
open-pdf-uri
pdf-page?
pdf-page
3 Rendering
page->pict
page->bitmap
page-render-to-dc!
page-render-to-cairo!
4 Layout
page-size
page-crop-box
page-text-in-rect
page-text
page-text-layout
page-text-with-layout
5 Searching
page-find-text
6 Metadata
pdf-count-pages
pdf-title
pdf-author
pdf-subject
pdf-keywords
pdf-creator
pdf-producer
page-label
7 Bugs and Issues

pdf-render: Render and parse PDF files

gcr, based on code by Jens Axel Søgaard

 (require (planet gcr/pdf-render:2:=0))
This library lets Racket render PDF files. You can also gather information about the text or layout.
This requires libpoppler, and thus likely only works on Linux or Mac OSX. Windows might work if you insctall Poppler, but I can’t make promises.

    1 Examples

    2 PDF files

    3 Rendering

    4 Layout

    5 Searching

    6 Metadata

    7 Bugs and Issues

1 Examples

Showing a PDF as a pict is as easy as you’d expect:
#lang racket
(require slideshow/pict
         (planet gcr/pdf-render))
 
(show-pict (page->pict "oopsla04-gff.pdf"))

By default, page->pict shows the first page of the given PDF filename. You can also say something like (show-pict (page->pict (pdf-page "oopsla04-gff.pdf" 5))) to show the 4th page (pages are zero-indexed).

You can also search for text. This snippet searches for every occurance of the word "the" and overlays a yellow rectangle on each:
;; The first page of a PDF file. (Pages are zero-indexed)
(define page (pdf-page "oopsla04-gff.pdf" 0))
 
;; Overlay each box over the PDF.
(for/fold ([pageview (page->pict page)])
   ([bounding-box (in-list (page-find-text page "the"))])
 (match-define (list x1 y1 x2 y2) bounding-box)
 ;; Each match's bounding box ^
 (pin-over pageview x1 y1
           (cellophane
            (colorize (filled-rectangle (- x2 x1) (- y2 y1)) "yellow")
            0.5)))

Finally, for low-level layout control, you can gather the bounding box geometry for each letter on the page. This snippet outlines each letter’s bounding box.
(define page (pdf-page "oopsla04-gff.pdf" 0))
(for/fold ([pageview (apply blank (page-size page))])
     ([box (in-list (page-text-layout page))])
   (match-define (list x1 y1 x2 y2) box)
   (pin-over pageview x1 y1
             (colorize (rectangle (- x2 x1) (- y2 y1))
                       "gray")))

Note that page->pict does not convert the file to a bitmap. You can rotate, scale, or transform the resulting pict however you like without losing quality.
(rotate
 (frame (scale (inset/clip (page->pict page) -400 -300 -100 -400) 5))
 (* 0.125 pi))

If you want to coerce the result to a bitmap for speed, you can do that as well:
(rotate
 (frame (scale (inset/clip (bitmap (page->bitmap page))
                           -400 -300 -100 -400)
               5))
 (* 0.125 pi))

2 PDF files

All functions that accept pages or documents also accept filenames. This is more convenient for you, but it is also less efficient because the document must be re-opened every time. You can make this faster by keeping the result of pdf-page or open-pdf-uri to ensure that this library only opens the document once.

(pdf-document? maybe-doc)  boolean?
  maybe-doc : any/c
Returns #t if maybe-doc is a PDF document. This can be a string (filename) or the result of open-pdf-uri.
(open-pdf-uri uri password)  (or/c pdf-document? false?)
  uri : string?
  password : (or/c string? false?)
Opens the given PDF file with the given uri and the given password. The password may be #f if the document does not have a password. The uri is usually something like "file:/tmp/test.pdf"; anything libpoppler accepts. Don’t forget the file: at the beginning for local documents!

This function will throw an error if the PDF file does not exist.

For example, if you wish to show the tenth page from an encrypted document, use:
(define document (open-pdf-uri "file:/tmp/secret.pdf" "some_password"))
(define page (pdf-page document 9))
(show-pict (page->pict page))
As a shortcut, (page->pict "/tmp/filename.pdf") is a shortcut for
(page->pict (pdf-page (open-pdf-uri "file:/tmp/filename.pdf" #f) 0))
Most of these functinos honor shortcuts like this.

(pdf-page? maybe-page)  boolean?
  maybe-page : any/c
Returns true if maybe-page is a page. Pages can be filenames (in which case the first page of the document will be used), pdf-document?s, or the result of pdf-page.

(pdf-page maybe-doc page-index)  pdf-page?
  maybe-doc : pdf-document?
  page-index : exact-nonnegative-integer?
Opens the page-indexth page in maybe-doc. Pages are zero-indexed, so if you want the third page of "/tmp/oopsla04-gff.pdf", for example, use:

(pdf-page "/tmp/oopsla04-gff.pdf" 2)

3 Rendering

(page->pict page)  pict?
  page : pdf-page?
Renders the given page to a pict from Racket’s slideshow/pict library. This pict can be transformed any way you like without rasterization.

(page->bitmap page)  (is-a?/c bitmap%)
  page : pdf-page?
Renders the page at 72 DPI to a bitmap%.

(page-render-to-dc! page dc)  any/c
  page : pdf-page?
  dc : (is-a?/c dc<%>)
Renders the given page to the given dc<%> context. This only works if dc is backed by Cairo and the moons are aligned properly.

(page-render-to-cairo! page _cairo_t)  any/c
  page : pdf-page?
  _cairo_t : any/c
Render the given page to the given _cairo_t Cairo context with the current transformation.

4 Layout

(page-size page)  
(list/c (and/c real? (not/c negative?))
        (and/c real? (not/c negative?)))
  page : pdf-page?
Returns the width and height of page, in points (1/72 of an inch).

(page-crop-box page)  
(list/c (and/c real? (not/c negative?))
        (and/c real? (not/c negative?))
        (and/c real? (not/c negative?))
        (and/c real? (not/c negative?)))
  page : pdf-page?
Returns the rectangle of page’s crop box.

Each rectangle is a (list x1 y1 x2 y2), where x1,y1 is the top left corner and x2,y2 is the bottom right. Coordinates are in points (1/72 of an inch).

(page-text-in-rect page mode x1 y1 x2 y2)  string?
  page : pdf-page?
  mode : (one-of/c 'glyph 'word 'line)
  x1 : (and/c inexact? (not/c negative?))
  y1 : (and/c inexact? (not/c negative?))
  x2 : (and/c inexact? (not/c negative?))
  y2 : (and/c inexact? (not/c negative?))
Returns the string inside the given rectangle. If mode is 'glyph, the closest-matching characters are returned; if mode is 'word, the returned string will be the words lying in the rectangle; if mode is 'line, an entire line of text will be returned.

When specifying the rectangle, x1,y1 should be the top left corner and x2,y2 should be the bottom right. Coordinates are in points (1/72 of an inch).

(page-text page)  string?
  page : pdf-page?
Returns all the text in page as a string.

(page-text-layout page)
  
(listof (list/c (and/c real? (not/c negative?))
                (and/c real? (not/c negative?))
                (and/c real? (not/c negative?))
                (and/c real? (not/c negative?))))
  page : pdf-page?
Returns bounding boxes for each of the letters in page. The poppler documentation says that the Nth character in (page-text page) should correspond with the Nth bounding box, but this may or may not be accurate.

Each bounding box is a (list x1 y1 x2 y2), where x1,y1 is the top left corner and x2,y2 is the bottom right. Coordinates are in points (1/72 of an inch).

(page-text-with-layout page)
  
(listof (list/c string
                (list/c (and/c real? (not/c negative?))
                        (and/c real? (not/c negative?))
                        (and/c real? (not/c negative?))
                        (and/c real? (not/c negative?)))))
  page : pdf-page?
Returns each letter in the document along with its bounding box for each of the letters in page. Note that each letter’s string may be longer than one letter, and it often includes newlines.

Each bounding box is a (list x1 y1 x2 y2), where x1,y1 is the top left corner and x2,y2 is the bottom right. Coordinates are in points (1/72 of an inch).

For example,
(take (page-text-with-layout "oopsla04-gff.pdf") 5)
will return the bounding boxes of the first five letters on the first page of "oopsla04-gff.pdf", namely:
'(("S\n" (150.738 71.302 162.699 87.890))
  ("u\n" (162.699 71.302 173.656 87.890))
  ("p\n" (173.656 71.302 184.613 87.890))
  ("e\n" (184.613 71.302 194.584 87.890))
  ("r\n" (194.584 71.302 201.560 87.890)))

5 Searching

(page-find-text page text)
  
(listof (list/c (and/c real? (not/c negative?))
                (and/c real? (not/c negative?))
                (and/c real? (not/c negative?))
                (and/c real? (not/c negative?))))
  page : pdf-page?
  text : string?
Searches for the given text on page. This is a case-insensitive search. Returns a list of rectangle coordinates. Each rectangle is a (list x1 y1 x2 y2), where x1,y1 is the top left corner and x2,y2 is the bottom right. Coordinates are in points (1/72 of an inch).

6 Metadata

Returns the number of pages in doc.

(pdf-title doc)  (or/c false? string?)
  doc : pdf-document?
(pdf-author doc)  (or/c false? string?)
  doc : pdf-document?
(pdf-subject doc)  (or/c false? string?)
  doc : pdf-document?
(pdf-keywords doc)  (or/c false? string?)
  doc : pdf-document?
(pdf-creator doc)  (or/c false? string?)
  doc : pdf-document?
(pdf-producer doc)  (or/c false? string?)
  doc : pdf-document?
Returns metadata about doc.

(page-label page)  (or/c false? string?)
  page : pdf-page?
Returns the page’s "label", which is often a numeral like "1" or "2", but it may be a Roman numeral like "xii" for certain sections of the PDF. It may even be some other exotic name like "Table\nof Contents".

7 Bugs and Issues

Note that pdf->pict draws directly to the underlying surface’s cairo context. This may have problems if the dc<%> is not backed by cairo or if you perform different transformations (like cropping or blurring).