Recipes: Text

How to Extract all Document Text

This script will take a document filename and generate a text file from all of its text.

The document can be any supported type like PDF, XPS, etc.

The script works as a command line tool which expects the document filename supplied as a parameter. It generates one text file named “filename.txt” in the script directory. Text of pages is separated by a form feed character:

import sys, fitz
fname = sys.argv[1]  # get document filename
doc = fitz.open(fname)  # open document
out = open(fname + ".txt", "wb")  # open text output
for page in doc:  # iterate the document pages
    text = page.get_text().encode("utf8")  # get plain text (is in UTF-8)
    out.write(text)  # write text of page
    out.write(bytes((12,)))  # write page delimiter (form feed 0x0C)
out.close()

The output will be plain text as it is coded in the document. No effort is made to prettify in any way. Specifically for PDF, this may mean output not in usual reading order, unexpected line breaks and so forth.

You have many options to rectify this – see chapter Appendix 2: Considerations on Embedded Files. Among them are:

  1. Extract text in HTML format and store it as a HTML document, so it can be viewed in any browser.

  2. Extract text as a list of text blocks via Page.get_text(“blocks”). Each item of this list contains position information for its text, which can be used to establish a convenient reading order.

  3. Extract a list of single words via Page.get_text(“words”). Its items are words with position information. Use it to determine text contained in a given rectangle – see next section.

See the following two sections for examples and further explanations.

How to Extract Text from within a Rectangle

There is now (v1.18.0) more than one way to achieve this. We therefore have created a folder in the PyMuPDF-Utilities repository specifically dealing with this topic.


How to Extract Text in Natural Reading Order

One of the common issues with PDF text extraction is, that text may not appear in any particular reading order.

This is the responsibility of the PDF creator (software or a human). For example, page headers may have been inserted in a separate step – after the document had been produced. In such a case, the header text will appear at the end of a page text extraction (although it will be correctly shown by PDF viewer software). For example, the following snippet will add some header and footer lines to an existing PDF:

doc = fitz.open("some.pdf")
header = "Header"  # text in header
footer = "Page %i of %i"  # text in footer
for page in doc:
    page.insert_text((50, 50), header)  # insert header
    page.insert_text(  # insert footer 50 points above page bottom
        (50, page.rect.height - 50),
        footer % (page.number + 1, doc.page_count),
    )

The text sequence extracted from a page modified in this way will look like this:

  1. original text

  2. header line

  3. footer line

PyMuPDF has several means to re-establish some reading sequence or even to re-generate a layout close to the original:

  1. Use sort parameter of Page.get_text(). It will sort the output from top-left to bottom-right (ignored for XHTML, HTML and XML output).

  2. Use the fitz module in CLI: python -m fitz gettext ..., which produces a text file where text has been re-arranged in layout-preserving mode. Many options are available to control the output.

You can also use the above mentioned script with your modifications.


How to Extract Tables from Documents

If you see a table in a document, you are not normally looking at something like an embedded Excel or other identifiable object. It usually is just text, formatted to appear as appropriate.

Extracting a tabular data from such a page area therefore means that you must find a way to (1) graphically indicate table and column borders, and (2) then extract text based on this information.

The wxPython GUI script wxTableExtract.py strives to exactly do that. You may want to have a look at it and adjust it to your liking.


How to Mark Extracted Text

There is a standard search function to search for arbitrary text on a page: Page.search_for(). It returns a list of Rect objects which surround a found occurrence. These rectangles can for example be used to automatically insert annotations which visibly mark the found text.

This method has advantages and drawbacks. Pros are:

  • The search string can contain blanks and wrap across lines

  • Upper or lower case characters are treated equal

  • Word hyphenation at line ends is detected and resolved

  • Return may also be a list of Quad objects to precisely locate text that is not parallel to either axis – using Quad output is also recommend, when page rotation is not zero

But you also have other options:

import sys
import fitz

def mark_word(page, text):
    """Underline each word that contains 'text'.
    """
    found = 0
    wlist = page.getTex("words")  # make the word list
    for w in wlist:  # scan through all words on page
        if text in w[4]:  # w[4] is the word's string
            found += 1  # count
            r = fitz.Rect(w[:4])  # make rect from word bbox
            page.add_underline_annot(r)  # underline
    return found

fname = sys.argv[1]  # filename
text = sys.argv[2]  # search string
doc = fitz.open(fname)

print("underlining words containing '%s' in document '%s'" % (word, doc.name))

new_doc = False  # indicator if anything found at all

for page in doc:  # scan through the pages
    found = mark_word(page, text)  # mark the page's words
    if found:  # if anything found ...
        new_doc = True
        print("found '%s' %i times on page %i" % (text, found, page.number + 1))

if new_doc:
    doc.save("marked-" + doc.name)

This script uses Page.get_text("words")() to look for a string, handed in via cli parameter. This method separates a page’s text into “words” using spaces and line breaks as delimiters. Further remarks:

  • If found, the complete word containing the string is marked (underlined) – not only the search string.

  • The search string may not contain spaces or other white space.

  • As shown here, upper / lower cases are respected. But this can be changed by using the string method lower() (or even regular expressions) in function mark_word.

  • There is no upper limit: all occurrences will be detected.

  • You can use anything to mark the word: ‘Underline’, ‘Highlight’, ‘StrikeThrough’ or ‘Square’ annotations, etc.

  • Here is an example snippet of a page of this manual, where “MuPDF” has been used as the search string. Note that all strings containing “MuPDF” have been completely underlined (not just the search string).

_images/img-markedpdf.jpg

How to Mark Searched Text

This script searches for text and marks it:

# -*- coding: utf-8 -*-
import fitz

# the document to annotate
doc = fitz.open("tilted-text.pdf")

# the text to be marked
t = "¡La práctica hace el campeón!"

# work with first page only
page = doc[0]

# get list of text locations
# we use "quads", not rectangles because text may be tilted!
rl = page.search_for(t, quads = True)

# mark all found quads with one annotation
page.add_squiggly_annot(rl)

# save to a new PDF
doc.save("a-squiggly.pdf")

The result looks like this:

_images/img-textmarker.jpg

How to Mark Non-horizontal Text

The previous section already shows an example for marking non-horizontal text, that was detected by text searching.

But text extraction with the “dict” / “rawdict” options of Page.get_text() may also return text with a non-zero angle to the x-axis. This is indicated by the value of the line dictionary’s "dir" key: it is the tuple (cosine, sine) for that angle. If line["dir"] != (1, 0), then the text of all its spans is rotated by (the same) angle != 0.

The “bboxes” returned by the method however are rectangles only – not quads. So, to mark span text correctly, its quad must be recovered from the data contained in the line and span dictionary. Do this with the following utility function (new in v1.18.9):

span_quad = fitz.recover_quad(line["dir"], span)
annot = page.add_highlight_annot(span_quad)  # this will mark the complete span text

If you want to mark the complete line or a subset of its spans in one go, use the following snippet (works for v1.18.10 or later):

line_quad = fitz.recover_line_quad(line, spans=line["spans"][1:-1])
page.add_highlight_annot(line_quad)
_images/img-linequad.jpg

The spans argument above may specify any sub-list of line["spans"]. In the example above, the second to second-to-last span are marked. If omitted, the complete line is taken.


How to Analyze Font Characteristics

To analyze the characteristics of text in a PDF use this elementary script as a starting point:

import fitz


def flags_decomposer(flags):
    """Make font flags human readable."""
    l = []
    if flags & 2 ** 0:
        l.append("superscript")
    if flags & 2 ** 1:
        l.append("italic")
    if flags & 2 ** 2:
        l.append("serifed")
    else:
        l.append("sans")
    if flags & 2 ** 3:
        l.append("monospaced")
    else:
        l.append("proportional")
    if flags & 2 ** 4:
        l.append("bold")
    return ", ".join(l)


doc = fitz.open("text-tester.pdf")
page = doc[0]

# read page text as a dictionary, suppressing extra spaces in CJK fonts
blocks = page.get_text("dict", flags=11)["blocks"]
for b in blocks:  # iterate through the text blocks
    for l in b["lines"]:  # iterate through the text lines
        for s in l["spans"]:  # iterate through the text spans
            print("")
            font_properties = "Font: '%s' (%s), size %g, color #%06x" % (
                s["font"],  # font name
                flags_decomposer(s["flags"]),  # readable font flags
                s["size"],  # font size
                s["color"],  # font color
            )
            print("Text: '%s'" % s["text"])  # simple print of text
            print(font_properties)

Here is the PDF page and the script output:

_images/img-pdftext.jpg

How to Insert Text

PyMuPDF provides ways to insert text on new or existing PDF pages with the following features:

  • choose the font, including built-in fonts and fonts that are available as files

  • choose text characteristics like bold, italic, font size, font color, etc.

  • position the text in multiple ways:

    • either as simple line-oriented output starting at a certain point,

    • or fitting text in a box provided as a rectangle, in which case text alignment choices are also available,

    • choose whether text should be put in foreground (overlay existing content),

    • all text can be arbitrarily “morphed”, i.e. its appearance can be changed via a Matrix, to achieve effects like scaling, shearing or mirroring,

    • independently from morphing and in addition to that, text can be rotated by integer multiples of 90 degrees.

All of the above is provided by three basic Page, resp. Shape methods:

Note

Both text insertion methods automatically install the font as necessary.

How to Write Text Lines

Output some text lines on a page:

import fitz
doc = fitz.open(...)  # new or existing PDF
page = doc.new_page()  # new or existing page via doc[n]
p = fitz.Point(50, 72)  # start point of 1st line

text = "Some text,\nspread across\nseveral lines."
# the same result is achievable by
# text = ["Some text", "spread across", "several lines."]

rc = page.insert_text(p,  # bottom-left of 1st char
                     text,  # the text (honors '\n')
                     fontname = "helv",  # the default font
                     fontsize = 11,  # the default font size
                     rotate = 0,  # also available: 90, 180, 270
                     )
print("%i lines printed on page %i." % (rc, page.number))

doc.save("text.pdf")

With this method, only the number of lines will be controlled to not go beyond page height. Surplus lines will not be written and the number of actual lines will be returned. The calculation uses a line height calculated from the fontsize and 36 points (0.5 inches) as bottom margin.

Line width is ignored. The surplus part of a line will simply be invisible.

However, for built-in fonts there are ways to calculate the line width beforehand - see get_text_length().

Here is another example. It inserts 4 text strings using the four different rotation options, and thereby explains, how the text insertion point must be chosen to achieve the desired result:

import fitz
doc = fitz.open()
page = doc.new_page()
# the text strings, each having 3 lines
text1 = "rotate=0\nLine 2\nLine 3"
text2 = "rotate=90\nLine 2\nLine 3"
text3 = "rotate=-90\nLine 2\nLine 3"
text4 = "rotate=180\nLine 2\nLine 3"
red = (1, 0, 0) # the color for the red dots
# the insertion points, each with a 25 pix distance from the corners
p1 = fitz.Point(25, 25)
p2 = fitz.Point(page.rect.width - 25, 25)
p3 = fitz.Point(25, page.rect.height - 25)
p4 = fitz.Point(page.rect.width - 25, page.rect.height - 25)
# create a Shape to draw on
shape = page.new_shape()

# draw the insertion points as red, filled dots
shape.draw_circle(p1,1)
shape.draw_circle(p2,1)
shape.draw_circle(p3,1)
shape.draw_circle(p4,1)
shape.finish(width=0.3, color=red, fill=red)

# insert the text strings
shape.insert_text(p1, text1)
shape.insert_text(p3, text2, rotate=90)
shape.insert_text(p2, text3, rotate=-90)
shape.insert_text(p4, text4, rotate=180)

# store our work to the page
shape.commit()
doc.save(...)

This is the result:

_images/img-inserttext.jpg

How to Fill a Text Box

This script fills 4 different rectangles with text, each time choosing a different rotation value:

import fitz
doc = fitz.open(...)  # new or existing PDF
page = doc.new_page()  # new page, or choose doc[n]
r1 = fitz.Rect(50,100,100,150)  # a 50x50 rectangle
disp = fitz.Rect(55, 0, 55, 0)  # add this to get more rects
r2 = r1 + disp  # 2nd rect
r3 = r1 + disp * 2  # 3rd rect
r4 = r1 + disp * 3  # 4th rect
t1 = "text with rotate = 0."  # the texts we will put in
t2 = "text with rotate = 90."
t3 = "text with rotate = -90."
t4 = "text with rotate = 180."
red  = (1,0,0)  # some colors
gold = (1,1,0)
blue = (0,0,1)
"""We use a Shape object (something like a canvas) to output the text and
the rectangles surrounding it for demonstration.
"""
shape = page.new_shape()  # create Shape
shape.draw_rect(r1)  # draw rectangles
shape.draw_rect(r2)  # giving them
shape.draw_rect(r3)  # a yellow background
shape.draw_rect(r4)  # and a red border
shape.finish(width = 0.3, color = red, fill = gold)
# Now insert text in the rectangles. Font "Helvetica" will be used
# by default. A return code rc < 0 indicates insufficient space (not checked here).
rc = shape.insert_textbox(r1, t1, color = blue)
rc = shape.insert_textbox(r2, t2, color = blue, rotate = 90)
rc = shape.insert_textbox(r3, t3, color = blue, rotate = -90)
rc = shape.insert_textbox(r4, t4, color = blue, rotate = 180)
shape.commit()  # write all stuff to page /Contents
doc.save("...")

Several default values were used above: font “Helvetica”, font size 11 and text alignment “left”. The result will look like this:

_images/img-textbox.jpg

How to Use Non-Standard Encoding

Since v1.14, MuPDF allows Greek and Russian encoding variants for the Base14_Fonts. In PyMuPDF this is supported via an additional encoding argument. Effectively, this is relevant for Helvetica, Times-Roman and Courier (and their bold / italic forms) and characters outside the ASCII code range only. Elsewhere, the argument is ignored. Here is how to request Russian encoding with the standard font Helvetica:

page.insert_text(point, russian_text, encoding=fitz.TEXT_ENCODING_CYRILLIC)

The valid encoding values are TEXT_ENCODING_LATIN (0), TEXT_ENCODING_GREEK (1), and TEXT_ENCODING_CYRILLIC (2, Russian) with Latin being the default. Encoding can be specified by all relevant font and text insertion methods.

By the above statement, the fontname helv is automatically connected to the Russian font variant of Helvetica. Any subsequent text insertion with this fontname will use the Russian Helvetica encoding.

If you change the fontname just slightly, you can also achieve an encoding “mixture” for the same base font on the same page:

import fitz
doc=fitz.open()
page = doc.new_page()
shape = page.new_shape()
t="Sômé tèxt wìth nöñ-Lâtîn characterß."
shape.insert_text((50,70), t, fontname="helv", encoding=fitz.TEXT_ENCODING_LATIN)
shape.insert_text((50,90), t, fontname="HElv", encoding=fitz.TEXT_ENCODING_GREEK)
shape.insert_text((50,110), t, fontname="HELV", encoding=fitz.TEXT_ENCODING_CYRILLIC)
shape.commit()
doc.save("t.pdf")

The result:

_images/img-encoding.jpg

The snippet above indeed leads to three different copies of the Helvetica font in the PDF. Each copy is uniquely identified (and referenceable) by using the correct upper-lower case spelling of the reserved word “helv”:

for f in doc.get_page_fonts(0): print(f)

[6, 'n/a', 'Type1', 'Helvetica', 'helv', 'WinAnsiEncoding']
[7, 'n/a', 'Type1', 'Helvetica', 'HElv', 'WinAnsiEncoding']
[8, 'n/a', 'Type1', 'Helvetica', 'HELV', 'WinAnsiEncoding']