when text that has associated regions is deleted

[xeno.by]

...then the regions are not deleted, but rather fall through (e.g. if I had a file with 10 lines and a region on the 5th line, then after deleting the 5th line, the region is now associated with the former 6th line). Is this a bug or a feature?

[jngeist]

I'm dealing with a related issue, and here's how I understand it: this is neither a bug nor a feature, but a misunderstanding of the metaphor. There's no such thing as "text that has associated regions." You can use view.add_regions (which is what I assume you're talking about) to add a set of *regions* to a view under a particular key, but that has nothing to do with *text.* Regions are locations in the file.

Let's say I have a buffer that just contains the text "foobar". I'm doing the following in the console:

Code: Select all


>>> a = sublime.Region(0, 3)
>>> a
(0, 3)
# Note that this isn't text at all. Just locations.

>>> view.substr(a)
u'foo'
# I can use those locations to get the text occupying that space *from the view.*

# Now, if I delete "foo", so that the buffer just contains "bar":
>>> view.substr(a)
u'bar'
>>> a
(0, 3)
# a (the Region) hasn't changed, but the contents of that Region have.


It's like the difference between addresses and people. view.add_regions allows you to set up a list of addresses, just like if you had a mailing list. But if a family moves, their old house stays at the same address, and so sending a letter to 123 Whatever Blvd will still go to the same place, whether they live there or not.

So: Regions are *supposed* to be independent of text, and so shouldn't be changed if the text is.

I think what we're both wanting, though, is to have something like mail forwarding (metaphorically, not programmatically), such that we can identify a span of text in a buffer with which we can associate data that will persist even if the location of that text changes. I can see, though, how this could be a programming nightmare, and I'm not sure at all how I'd go about implementing it myself. I'm hoping there's something in place that can simulate that, though. Maybe?

[xeno.by]

Yep it seems so. Great metaphor by the way!

I decided to live with it. In my use case regions denote breakpoints (to provide an indicator that certain line has a breakpoint associated with it), so I don't trust add_regions and keep all the info in an external settings file. Then in an on_modified event of a sublime.EventListener I check whether recent edits skewed the regions and fix them up if necessary.