Skip to content

Instantly share code, notes, and snippets.

Embed
What would you like to do?
pdf_to_html_preview messy hack to work on Windows
#forked from: Julian_Todd / PDF to HTML (https://scraperwiki.com/views/pdf-to-html-preview-1/)
#input url goes to line
import sys
import urllib, urllib2, urlparse
import lxml.etree, lxml.html
import re, os
def Pageblock(page, index):
'''
Print each page of the PDF in turn, outputting the contents as HTML.
'''
result = [ ]
assert page.tag == 'page'
height = int(page.attrib.get('height'))
width = int(page.attrib.get('width'))
number = page.attrib.get('number')
assert page.attrib.get('position') == "absolute"
result.append('<p>Page %s index %d height=%d width=%d</p>' % (number, index, height, width))
result.append('<div class="page" style="height:%dpx; width:%dpx">' % (height, width))
for v in page:
if v.tag == 'fontspec':
continue
assert v.tag == 'text'
text = re.match('(?s)<text.*?>(.*?)</text>', lxml.etree.tostring(v)).group(1)
top = int(v.attrib.get('top'))
left = int(v.attrib.get('left'))
width = int(v.attrib.get('width'))
height = int(v.attrib.get('height'))
fontid = v.attrib.get('font')
style = 'top:%dpx; left:%dpx; height:%dpx; width:%dpx' % (top, left, height, width)
result.append(' <div class="text fontspec-%s" style="%s">%s</div>' % (fontid, style, text))
result.append('</div>')
return '\n'.join(result)
# pdftoxml based on utils.py in scraperwiki-python
# David Jones, ScraperWiki Limited
# Thomas Levine, ScraperWiki Limited
# BSD Licence; see https://github.com/scraperwiki/scraperwiki-python
def pdftoxml(pdfdata, options):
"""converts pdf file to xml file"""
# lots of hacky Windows fixes c.f. original
with open('input.pdf', 'wb') as f:
f.write(pdfdata)
cmd = 'pdftohtml -xml -nodrm -zoom 1.5 -enc UTF-8 -noframes '
if options:
cmd += options
cmd += 'input.pdf output.xml'
cmd = cmd + " > NUL 2>&1" # can't turn off output, so throw away even stderr yeuch
os.system(cmd)
with open('output.xml', 'r') as f:
return f.read()
def Main(pdfurl, hidden):
'''
Take the URL of a PDF, and use scraperwiki.pdftoxml and lxml to output the contents
as a styled HTML div.
'''
pdfdata = urllib2.urlopen(pdfurl).read()
options = ''
if hidden == 'hidden':
options='-hidden ' #
# TODO: readd this if implemented in scraperwiki-python
# see https://github.com/scraperwiki/scraperwiki-python/issues/48
# pdfxml = scraperwiki.pdftoxml(pdfdata, options)
pdfxml = pdftoxml(pdfdata, options)
try:
root = lxml.etree.fromstring(pdfxml)
except lxml.etree.XMLSyntaxError, e:
print str(e), str(type(e)).replace("<", "&lt;")
print pdfurl
print pdfxml.replace("<", "&lt;")
root = []
global styles
fontspecs = { }
# Get the PDF's internal styles: we'll use these to style the divs containing the PDF.
for fontspec in (root is not None and root.xpath('page/fontspec')):
id = fontspec.attrib.get('id')
fontdesc = {'size':int(fontspec.attrib.get('size')), 'family':fontspec.attrib.get('family'), 'color':fontspec.attrib.get('color')}
fontspecs[id] = fontdesc
styles['div.fontspec-%s' % id] = 'color:%s;font-family:%s;font-size:%dpx' % (fontdesc['color'], fontdesc['family'], fontdesc['size'])
# Output the view, with instructions for the user.
print '<html dir="ltr" lang="en">'
print '<head>'
print ' <meta http-equiv="Content-Type" content="text/html; charset=utf-8"/>'
print ' <title>PDF to XML text positioning</title>'
print ' <style type="text/css" media="screen">%s</style>' % "\n".join([ "%s { %s }" % (k, v) for k, v in styles.items() ])
print ' <script type="text/javascript" src="https://ajax.googleapis.com/ajax/libs/jquery/1.4.2/jquery.min.js"></script>'
print ' <script>%s</script>' % jscript
print '</head>'
print '<div class="info" id="info1">&lt;text block&gt;</div>'
print '<div class="info" id="info2">&lt;position&gt;</div>'
print '<div class="heading">'
print '<h2>Graphical preview of scraperwiki.pdftoxml(pdfdata)</h2>'
print '<p>Click on a text line to see its coordinates and any other text that shares the same column or row.'
print ' Useful for discovering what coordinates to use when extracting rows from tables in a document.</p>'
print '<p>To do: track the coordinates of the mouse and cross reference with <a href="/cropper">cropper</a> technology.</p>'
print '<p class="href"><a href="%s">%s</a></p>'% (pdfurl, pdfurl)
print '<form id="newpdfdoclink">'
print '<label for="url">PDF link</label>'
print ' <input type="text" name="url" id="url" value="%s" title="paste in url of document">' % pdfurl
if hidden == 1:
checked="checked "
else:
checked=""
print '<br /><label for="hidden">Force hidden text extraction</label>'
print ' <input type="checkbox" name="hidden" id="hidden" value="1" %stitle="force hidden text extraction">' % checked
print '<br /> <input type="submit" value="Go">'
print '</form>'
ttx = re.sub('<', '&lt;', pdfxml)
ttx = re.sub('\n', '\r\n', ttx)
# Does this truncate in case of large PDF?
print '<textarea class="pdfprev">%s</textarea>' % ttx[:5000]
print '</div>'
print '<p>There are %d pages</p>' % len(root)
# Print each page of the PDF.
for index, page in enumerate(root):
print Pageblock(page, index)
# Global styles for the divs containing the PDF.
styles = {
"div#info1": "position:fixed; white-space:pre; background-color:#ffd; border: thin red solid; z-index: 50; top:0px;",
"div#info2": "position:fixed; white-space:pre; background-color:#ffd; border: thin red solid; z-index: 50; top:20px;",
"div.heading": "padding-left:150px;",
"p.href": "font-size:60%",
"div.page": "background-color:#fff; border:thin black solid; position:relative; margin:2em;",
"div.text": "position:absolute; white-space:pre; background-color:#eee;",
"textarea.pdfprev":"white-space:pre; height:150px; width:80%",
"div.text:hover": "background-color:#faa; cursor:pointer",
"div.linev": "background-color:#fcc;",
"div.lineh": "background-color:#fce;",
}
# Global JavaScript allowing the user to click on an area of the PDF div, and see the
# underlying PDF source.
jscript = """
var rfontspec = new RegExp('fontspec-(\\\\w+)');
$(function()
{
$('div.text').click(function ()
{
var top = parseInt($(this).css('top'));
var left = parseInt($(this).css('left'));
var width = parseInt($(this).css('width'));
var height = parseInt($(this).css('height'));
var clas = $(this).attr('class');
var lfont = rfontspec.exec(clas);
var font = (lfont ? lfont[1] : clas);
$('div#info1').text($(this).html());
$('div#info2').text('top='+top + ' bottom='+(top+height)+ ' left='+left + ' right='+(left+width) + ' font='+font);
$('div.text').each(function()
{
var lleft = parseInt($(this).css('left'));
if (lleft == left)
$(this).addClass('linev');
else
$(this).removeClass('linev');
var ltop = parseInt($(this).css('top'));
if (ltop == top)
$(this).addClass('lineh');
else
$(this).removeClass('lineh');
});
});
});
"""
# Check for a PDF URL and hidden-checkbox entered by the user: if none, use our default values:
# urlquery = os.getenv('URLQUERY')
#if urlquery:
# querydata = urlparse.parse_qsl(urlquery);
# for pair in querydata:
# if pair[0] == "url":
# pdfurl = urllib.unquote(pair[1])
# if pair[0] == "hidden":
# hidden = 1
if __name__ == '__main__':
try:
pdfurl = sys.argv[1]
except IndexError:
print "Enter a valid URL. For example:"
print "http://soswy.state.wy.us/Elections/Docs/2004/04Results/04General/AL_PbP_Candidate_Summary.pdf"
sys.exit(1)
try:
hidden = sys.argv[2]
except IndexError:
hidden = None
Main(pdfurl, hidden)
@StevenMaude

This comment has been minimized.

Copy link
Owner Author

commented May 16, 2014

Windows installation

Python

Install the latest Python 2 (this is 2.7.6 at the moment). This was all tested with the x86 release, so I'd recommend that (not x86-64).

Make a directory to work in, e.g. pdf_to_html

Open command prompt in this work folder (quick way, in Explorer, right click the folder while holding shift, Open Command Prompt).

Try running Python by entering python into the command prompt. If you get 'python' is not recognized..., execute this command: set PATH=%PATH%;"C:\Python27" (That's the default for Python in Windows; otherwise replace the Python directory with wherever your Python lives.)

If you need this change, it only persists in this command prompt; there are plenty of guides on how to set this permanently using Windows Control Panel.

At this stage, you should be able to execute the command python at the prompt, and you'll be in a Python interpreter. Exit this by entering exit() to return to the command prompt.

lxml

Download lxml from https://pypi.python.org/pypi/lxml/3.3.5

Get lxml-3.3.5.win32-py2.7.exe

You can install lxml. Alternatively, you can just open the archive up with e.g. 7-Zip (it's an executable zip file) and extract the lxml directory to your work pdf_to_html folder

Start the Python interpreter again in the same command prompt with python; enter import lxml. If all is well, you should see nothing, it should execute silently (without any ImportError). Great! Exit back to the command prompt again with exit().

pdftohtml.exe

We need a recent version of the pdftohtml program that does the conversion from PDF to XML. An easy way to get this on Windows is via Calibre Portable. Install it anywhere; you just need a few files from it. From where you installed it, from the folder containing calibre.exe, copy pdftohtml.exe into your pdf_to_html work folder. Next, go into the DLLs folder in the Calibre install, copy freetype.dll, jpeg.dll, libpng12.dll, zlib1.dll to your pdf_to_html directory.

This pdf_to_html.py script

Download the pdf_to_html.py from this gist and save as pdf_to_html.py; this is hacked to work on Windows without needing the scraperwiki package. Move this to your work pdf_to_html folder.

Usage

Right, after all that hard work, you should have pdf_to_html.py, pdftohtml.exe and the DLLs from Calibre with the lxml folder all in your pdf_to_html folder.

Execute in the command prompt you've been using this whole time with python pdf_to_html.py http://www.somepdf.url > output.html

If you want to use the hidden text option, python pdf_to_html.py http://www.somepdf.url hidden > output.html

View output.html in a browser to see the output.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.