[kovidgoyal]

Originally Posted by minstrel

My overall plan is to write a Calibre plugin that takes the HTML input on importing and does some very basic, regular-expression based substitution on the file so some Browser specific tags can be kicked out, and possibly some CSS-constructs not supported by epub may be replaced. I have seen similar ideas floating around on this forum, but nobody (yet) seems to have tried to address it in the form of a Calibre plugin.

That's a good idea, if you come up with a reasonably general plugin, I'll be happy to distribute it with calibre.

My results (about three hours of playing around with writing plugins) have been mixed. Here are my findings so far (maybe the first three should go into the documentation):

1) The plugin name must end with _plugin.py, otherwise Calibre won't find it in the zip container.

Yeah, the documentation does specify that this is the case.

2) When writing plugins on a windows machine I had to be sure to save the .py file in UNIX format. Windows/DOS format will not work, when adding such a plugin to Calibre it will choke right on line 1 with a weird error.

Hmm calibre just executes the plugin as a normal python module using the python interpreter, so this shouldn't be a problem, but I'll implement a fix for it in the next release.

3) The Hello World example differs between the example given on the web page and the one available as a file download (from the same page):

Oops, Now fixed.

5) Why does the example use a FileTypePlugin when it actually modifies the Metadata? Shouldn't it rather be a Metadata Plugin?

Creating a metadata plugin would mean that the publisher is overridden in *all* contexts, whereas using a filetype plugin you can ensure that it is only overriden on import or preprocess or whatever.

6) I tried to write a simple plugin which simply replaces the word "Hello" by "World". I can nicely install it in Calibre, alas, it won't do anything. Is my approach completely wrong (do I have to do some XML tree processing, do I have to do something special to get to the content, is the temporary_file() method used correctly?):

You shouldn't have to, but the case of HTML import is a little complicated because the HTML file is transformed into a ZIP file on import by another plugin, so perhaps the two plugins are interacting to cause this problem. I'll look into it in a little bit.

As a general note you should probably change your plugin to run on preprocess rather than on import, Since people dont always use the calibre GUI to do conversions.