TP wrote:

It seems not possible to define with fromstring() or ET.XML a tree containing attributes with spaces.

I do hope it isn't.

But it is possible by adding the attribute containing a space afterwards, see the example below.

################### #!/usr/bin/env python # -*- coding: utf-8 -*-

import lxml.etree as ET

root = ET.XML("<root><foo attri='bar'>data</foo></root>") foo_elem = root.xpath( "//foo" ) foo_elem[0].set( "tu tu", "22" ) print ET.tostring( root ) ###################

We obtain: <root><foo attri="bar" tu tu="22">data</foo></root>

Hmmm, ok, that looks like a bug to me. lxml should validate attribute names on the way in, just like tag names are validated.

From another point of view, often we would like to define attribute names as they are, i.e. english expressions with spaces.

How do you know that they will only ever be "english" expressions? What about Farsi and Chinese?

How do you proceed? Put underscores in the attribute names, and then remove them when displaying in the tree (for example in a graphical widget)?

It is a very good and common design choice to separate data from representation. So these two are completely orthogonal. You can use '_' or '-' to separate words, or you can use a prefixed MD5 hash for the attribute name that maps to a separate name lookup table. Choices are endless.

Or define the correspondance between the attribute names and the english names in some part of the XML file (for example, the attribute names could be tags, associated to some text that would contain the english names.

With "tags" you mean "references", I assume. Maybe even references into a separate XML file (one per language) that defines the presentational name. Without knowing enough about your application, this sounds like a reasonable thing to do. Stefan