If I return the x position of an object with the query-x action, then use the returned $value in a new command with the action object-set-attribute:x,$value, the object moves!

Is there any way to make sure these commands use the same units regardless of the document properties?

The query series - from the command line return the visual bounding box in pixel. 

From an extension you would use:

# Unit conversions
conversions = {
    'in': 96.0,
    'pt': 1.3333333333333333,
    'px': 1.0,
    'mm': 3.779527559055118,
    'cm': 37.79527559055118,
    'm': 3779.527559055118,
    'km': 3779527.559055118,
    'Q': 0.94488188976378,
    'pc': 16.0,
    'yd': 3456.0,
    'ft': 1152.0,
    '': 1.0,  # Default px
}

and then conversion_factor = conversions[self.svg.unit]

from the command line you would probably be better using inkex with python ( it does not require inkscape ) 

or you could parse the document with some other xml aware program - you would still use the conversion table

Thanks for your answer.

Do you know which unit is each action using between query-x and set-attribute?

I'm not writing a python extension but a bash script and would like to keep it in one file.

set-attribute writes directly to the attribute

so If you just set 20 for example that has no units. 

However I would try to parse the file in that case to find the units. 

this was the best I was able to do - to return units from an Inkscape .svg file.

xmlstarlet sel -t -v "//*['sodipodi:namedview']/@inkscape:document-units" drawing.svg

Thank you!

Dividing the queried x by the document unit and feeding the result to set-attribute works, although if there is a page offset (since multipage svgs were introduced), it has to be added to the result in svg unit (i.e. no unit).

I also found out that the transform-translate action uses the same unit as query, which is great because I actually want to perform translations.

Problem double-solved for me.

I spoke too fast. Turns out I can't use the query function because it's not just a unit problem. For example, the position of a text object's bounding box will differ depending on the text content, so in my case I need the svg value for x.

ChatGPT showed me this command which seems to work fine, where text453 is my object's id:

xmlstarlet sel -t -v "//*[@id='text453']/@x" drawing.svg

I've never used chatgpt - what did you use as a question ? 

Here's the prompt I used in chatGPT:

Using xmlstarlet on an svg file, how can I query the x position of an object of which I know only the id?