lxml使用元素移动文本

Question

I have an issue with wrapping image with a div.

from lxml.html import fromstring
from lxml import etree

tree = fromstring('<img src="/img.png"/> some text')
div = etree.Element('div')
div.insert(0, tree.find('img'))
tree.insert(0, div)
print etree.tostring(tree)

<span><div><img src="/img.png"/> some text</div></span>

Why does it add a span and how can I make it wrap image without text?

Answer 1

Because lxml is acutally an xml parser. It has some forgiving parsing rules that allows it to parse html (the lxml.html part), but it will internally always build a valid tree.

'<img src="/img.png"/> some text' isn't a tree, as it has no single root element, there is a img element, and a text node. To be able to store this snipplet internally, lxml needs to wrap it in a suitable tag. If you give it a string alone, it will wrap it in a p tag. Earlier versions just wrapped everything in html tags, which can lead to even more confusion.

You could also use html.fragment_fromstring , which doesn't add tags in that case, but would raise an error because the fragment isn't valid.

As for why the text sticks to the img tag: that's how lxml stores text. Take this example:

>>> p = html.fromstring("<p>spam<br />eggs</p>")
>>> br = p.find("br")
>>> p.text
'spam'
>>> br.text       # empty
>>> br.tail       # this is where text that comes after a tag is stored
'eggs'

So by moving a tag, you also move it's tail.

Answer 2

lxml.html is a kinder, gentler xml processor that tries to make sense of invalid xml. The sting you passed in is just junk from an xml perspective, but lxml.html wrapped it in a span element to make it valid again. If you don't want lxml.html guestimating, stick with lxml.etree.fromstring(). That version will reject the string.