[layout-] stop errors: hide-col on empty sheet, inputMultiple()

[midichef]

A minor fix for hide-col in version 2.11: if you accidentally press - too many times and hide on an empty sheet, you will see a scary exception: AttributeError: 'NoneType' object has no attribute 'hide'. This fixes that.

My logic for fixing these little error messages, is that it's good to keep new users from triggering them. When people are new and don't know how visidata works, and can't remember what all the keys do, it can be intimidating to explore. I remember thinking "Oh no, what did I just press, there is the red exclamation point again! Did I break something? Am I going to break it all worse?" I triggered this particular error myself by accident, so other people probably will too.

Should we add the same kind of protection for other keys that give errors on an empty sheet, like d and x?

The second fix is for the develop branch. If value is not provided to inputMultiple(), it will now fall back on a value of '' instead of None. As it is, if you run vd --debug to see the exceptions, and do search-col, you will see:

Traceback (most recent call last):
  File "search-col", line 1, in <module>
  File "/home/midichef/.local/lib/python3.10/site-packages/visidata/search.py", line 81, in moveInputRegex
    r = vd.inputMultiple(regex=dict(prompt=f"{action} regex: ", type=type, defaultLast=True),
  File "/home/midichef/.local/lib/python3.10/site-packages/visidata/_input.py", line 391, in inputMultiple
    promptlen = clipdraw(sheet._scr, y-v.get('dy'), promptlen, v.get('value'), attr, w=sheet.windowWidth-1)
  File "/home/midichef/.local/lib/python3.10/site-packages/visidata/cliptext.py", line 168, in clipdraw
    return clipdraw_chunks(scr, y, x, chunks, attr, w=w, clear=clear, rtl=rtl, **kwargs)
  File "/home/midichef/.local/lib/python3.10/site-packages/visidata/cliptext.py", line 194, in clipdraw_chunks
    for colorname, chunk in chunks:
  File "/home/midichef/.local/lib/python3.10/site-packages/visidata/cliptext.py", line 57, in iterchunks
    chunks = re.split(internal_markup_re, s)
  File "/usr/lib/python3.10/re.py", line 230, in split
    return _compile(pattern, flags).split(string, maxsplit)
TypeError: expected string or bytes-like object

[saulpw]

Yup, these make sense to me! We might also consider changing hide-col to fail() with a nice message, which would mean that replaying a cmdlog and hiding too many columns would abort the replay. Or we could just use warning() for the message alone. Do you think either of these would be preferable to the no-op?

[midichef]

Oh, good point, fail() seems better than the no-op. Aborting a replay is a useful sign that the data is in an unexpected state.

[frosencrantz]

My logic for fixing these little error messages, is that it's good to keep new users from triggering them. When people are new and don't know how visidata works, and can't remember what all the keys do, it can be intimidating to explore. I remember thinking "Oh no, what did I just press, there is the red exclamation point again! Did I break something? Am I going to break it all worse?" I triggered this particular error myself by accident, so other people probably will too.

I agree. Thank you for fixing these, @midichef! I think it is helpful to clean up these smaller issues, I feel it can make a difference in how new users feel about Visidata. If it doesn't look like it is easy to break, it will feel more robust. Even if the errors are not significant.

If you are looking at changing the message, it seems like it would be useful to mention the gv command to unhide all the columns or the C command to go to the columns sheet to change visibility.

[midichef]

If you are looking at changing the message, it seems like it would be useful to mention the gv command to unhide all the columns or the C command to go to the columns sheet to change visibility.

Oh, that's a good idea! It would be helpful.