Node.toprettyxml() adds newlines to DOCTYPE in Python

问题

When using prettify my DOCTYPE is broken into three lines. How can I keep it on one line?

The "broken" output:

<?xml version="1.0" encoding="utf-8"?>
<!DOCTYPE smil
  PUBLIC '-//W3C//DTD SMIL 2.0//EN'
  'http://www.w3.org/2001/SMIL20/SMIL20.dtd'>
<smil xmlns="http://www.w3.org/2001/SMIL20/Language">
  <head>
    <meta base="rtmp://cp23636.edgefcs.net/ondemand"/>
  </head>
  <body>
    <switch>
      <video src="mp4:soundcheck/1/clay_aiken/02_sc_ca_sorry_256.mp4" system-bitrate="336000"/>
      <video src="mp4:soundcheck/1/clay_aiken/02_sc_ca_sorry_512.mp4" system-bitrate="592000"/>
      <video src="mp4:soundcheck/1/clay_aiken/02_sc_ca_sorry_768.mp4" system-bitrate="848000"/>
      <video src="mp4:soundcheck/1/clay_aiken/02_sc_ca_sorry_1128.mp4" system-bitrate="1208000"/>
    </switch>
  </body>
</smil>

The script:

import csv
import sys
import os.path

from xml.etree import ElementTree
from xml.etree.ElementTree import Element, SubElement, Comment, tostring

from xml.dom import minidom

def prettify(doctype, elem):
    """Return a pretty-printed XML string for the Element.
    """
    rough_string = doctype + ElementTree.tostring(elem, 'utf-8')
    reparsed = minidom.parseString(rough_string)
    return reparsed.toprettyxml(indent="  ", encoding = 'utf-8')

doctype = '<!DOCTYPE smil PUBLIC "-//W3C//DTD SMIL 2.0//EN" "http://www.w3.org/2001/SMIL20/SMIL20.dtd">'

video_data = ((256, 336000),
              (512, 592000),
              (768, 848000),
              (1128, 1208000))


with open(sys.argv[1], 'rU') as f:
    reader = csv.DictReader(f)
    for row in reader:
        root = Element('smil')
        root.set('xmlns', 'http://www.w3.org/2001/SMIL20/Language')
        head = SubElement(root, 'head')
        meta = SubElement(head, 'meta base="rtmp://cp23636.edgefcs.net/ondemand"')
        body = SubElement(root, 'body')

        switch_tag = ElementTree.SubElement(body, 'switch')

        for suffix, bitrate in video_data:
            attrs = {'src': ("mp4:soundcheck/{year}/{id}/{file_root_name}_{suffix}.mp4"
                             .format(suffix=str(suffix), **row)),
                     'system-bitrate': str(bitrate),
                     }
            ElementTree.SubElement(switch_tag, 'video', attrs)

        file_root_name = row["file_root_name"]
        year = row["year"]
        id = row["id"]
        path = year+'-'+id

        file_name = row['file_root_name']+'.smil'
        full_path = os.path.join(path, file_name)
        output = open(full_path, 'w')
        output.write(prettify(doctype, root))

回答1:

Having looked over your current script and the other questions you've asked on this subject, I think you could make your life a lot simpler by building your smil files using string manipulation.

Almost all the xml in your files is static. The only data you need to worry about processing correctly is the attribute values for the video tag. And for that, there is a handy function in the standard library that does exactly what you want: xml.sax.saxutils.quoteattr.

So, with those points in mind, here is a script that should be a lot easier to work with:

import sys, os, csv
from xml.sax.saxutils import quoteattr

smil_header = '''\
<?xml version="1.0" encoding="utf-8"?>
<!DOCTYPE smil PUBLIC "-//W3C//DTD SMIL 2.0//EN" "http://www.w3.org/2001/SMIL20/SMIL20.dtd">
<smil xmlns="http://www.w3.org/2001/SMIL20/Language">
  <head>
    <meta base="rtmp://cp23636.edgefcs.net/ondemand"/>
  </head>
  <body>
    <switch>
'''
smil_video = '''\
      <video src=%s system-bitrate=%s/>
'''
smil_footer = '''\
    </switch>
  </body>
</smil>
'''

src_format = 'mp4:soundcheck/%(year)s/%(id)s/%(file_root_name)s_%(suffix)s.mp4'

video_data = (
    ('256', '336000'), ('512', '592000'),
    ('768', '848000'), ('1128', '1208000'),
    )

root = os.getcwd()
if len(sys.argv) > 2:
    root = sys.argv[2]

with open(sys.argv[1], 'rU') as stream:

    for row in csv.DictReader(stream):
        smil = [smil_header]
        for suffix, bitrate in video_data:
            row['suffix'] = suffix
            smil.append(smil_video % (
                quoteattr(src_format) % row, quoteattr(bitrate)
                ))
        smil.append(smil_footer)

        directory = os.path.join(root, '%(year)s-%(id)s' % row)
        try:
            os.makedirs(directory)
        except OSError:
            pass
        path = os.path.join(directory, '%(file_root_name)s.smil' % row)
        print ':: writing file:', path
        with open(path, 'wb') as stream:
            stream.write(''.join(smil))

回答2:

I think you have at least three options:

Just accept the newlines. They may be undesired and ugly but they are perfectly legal.

Add a kludge that replaces the bad DOCTYPE with a nicer one. Perhaps something like this:

import re

pretty_xml = prettify(doctype, elem)
m = re.search("(<!.*dtd'>)", pretty_xml, re.DOTALL)
ugly_doctype = m.group() 
fixed_xml = pretty_xml.replace(ugly_doctype, doctype)

Use a more feature-rich XML package. lxml springs to mind; it is mostly compatible with ElementTree. By using lxml's tostring function, you won't need the prettify function and the DOCTYPE comes out as you want it. Example:

from lxml import etree 

doctype = '<!DOCTYPE smil PUBLIC "-//W3C//DTD SMIL 2.0//EN" "http://www.w3.org/2001/SMIL20/SMIL20.dtd">'

XML = '<smil xmlns="http://www.w3.org/2001/SMIL20/Language"><head><meta base="rtmp://cp23636.edgefcs.net/ondemand"/></head><body><switch><video src="mp4:soundcheck/1/clay_aiken/02_sc_ca_sorry_256.mp4" system-bitrate="336000"/><video src="mp4:soundcheck/1/clay_aiken/02_sc_ca_sorry_512.mp4" system-bitrate="592000"/><video src="mp4:soundcheck/1/clay_aiken/02_sc_ca_sorry_768.mp4" system-bitrate="848000"/><video src="mp4:soundcheck/1/clay_aiken/02_sc_ca_sorry_1128.mp4" system-bitrate="1208000"/></switch></body></smil>'

elem = etree.fromstring(XML)
print etree.tostring(elem, doctype=doctype, pretty_print=True,
                     xml_declaration=True, encoding="utf-8")

Output:

<?xml version='1.0' encoding='utf-8'?>
<!DOCTYPE smil PUBLIC "-//W3C//DTD SMIL 2.0//EN" "http://www.w3.org/2001/SMIL20/SMIL20.dtd">
<smil xmlns="http://www.w3.org/2001/SMIL20/Language">
  <head>
    <meta base="rtmp://cp23636.edgefcs.net/ondemand"/>
  </head>
  <body>
    <switch>
      <video src="mp4:soundcheck/1/clay_aiken/02_sc_ca_sorry_256.mp4" system-bitrate="336000"/>
      <video src="mp4:soundcheck/1/clay_aiken/02_sc_ca_sorry_512.mp4" system-bitrate="592000"/>
      <video src="mp4:soundcheck/1/clay_aiken/02_sc_ca_sorry_768.mp4" system-bitrate="848000"/>
      <video src="mp4:soundcheck/1/clay_aiken/02_sc_ca_sorry_1128.mp4" system-bitrate="1208000"/>
    </switch>
  </body>
</smil>

回答3:

I don't believe it's possible to remove the newlines generated by Node.toprettyxml for the DOCTYPE, at least in a Pythonic way.

It's the writexml method of the DocumentType class, which starts at line 1284 of the minidom module, which inserts the offending newlines. The newline string inserted comes originally from the Node.toprettyxml method and is passed via the writexml method of the Document class. The same newline string is also passed to the writexml method of various other subclasses of Node. Changing the newline string in the call to Node.prettyxml will change the newline string used throughout the outputted XML.

There are various hacky ways around this: modify your local copy of the minidom module, 'monkey-patch' the writexml method of the DocumentType class or post-process the XML string to remove the unwanted newlines. However, none of these approaches appeal to me.

To me, the best approach seems to be to leave things as they are. Is it really a serious problem having the DOCTYPE split over multiple lines?

来源：https://stackoverflow.com/questions/8685032/node-toprettyxml-adds-newlines-to-doctype-in-python

标签

python

xml

pretty-print

elementtree

minidom