Can you build it without lxml?

[urain39]

MechanicalSoup is a really nice package i have used for, but it still requires C Compiler to compile the lxml on *nix systems.

It may be a problem to port to some platforms without C Compiler, such as Android or some minified Linux.

Currently i used a script to build MechanicalSoup without lxml:

#!/bin/sh

# Remove lxml in requirements.txt
sed -i '/lxml/d' requirements.txt

# Use `html.parser` instead `lxml`
sed -i "s@{'features': 'lxml'}@{'features': 'html.parser'}@g" mechanicalsoup/*.py

# Fix examples and tests
sed -i "s@\\(BeautifulSoup(.\\{1,\\}\\)'lxml'\\(.*)\\)@\1'html.parser'\2@g" examples/*.py tests/*.py

It works well, so i think it is not a big problem...

[moy]

lxml became the default in 83b2cdd, see the commit message for a bit more info. Essentially, the problem is:

• We really want a default value for the parser, because we want reproducibility (not having a default would mean that the same program using MechanicalSoup would behave differently depending on what is installed on the machine where it runs).
• lxml is rather clearly the best default in terms of features and performance, so although it adds a dependency we chose it as default. And therefore it has to be a dependency, otherwise the default configuration would be broken.
  I believe the simplest option for people who need to run MechanicalSoup on machines where a C compiler isn't installed is to patch it, as you did. We could also try to publish a MechanicalSoup-lite, or so, variant where lxml would not be a dependency but that would probably be a lot of trouble for a very small number of users. Up to now, I think we never got this request (except your message obviously).

[urain39]

@moy Thanks reply. I think the idea of releasing a lite version is good enough.