Add MiscTable.type

[hagenw]

When looping through all tables of a database it might become easier if MiscTable would have a few more properties from Table.
For example, if you would like to gather some metadata in a single loop over all tables, and you would like to store how many filewise tables you have, you would need a check like:

if not isinstance(table, audformat.MiscTable):
    if table.is_filewise:
        count += 1

First I thought it would be good to have

• is_filewise
• is_segmented
  and set both to False, but I we could then have the following inconsitency:

>>> table = audformat.MiscTable(pd.Index(["f1"], name="file", dtype="string"))
>>> table.is_filewise
False
>>> audformat.is_filewise_index(table)
True

So, I would maybe only thing about adding type and setting it to 'misc'?

[frankenjoe]

If you loop over db.tables you don't have to check if the table is a misc table.

[hagenw]

If you loop over db.tables you don't have to check if the table is a misc table.

Yes, but if you a lot of other stuff in the loop and some should affect tables and misc tables it becomes cumbersome to first loop tables and afterwards loop misc tables.

[frankenjoe]

So, I would maybe only thing about adding type and setting it to 'misc'?

Ok, so the idea is that your example above can be simplified to:

if table.type != 'misc':
    if table.is_filewise:
        count += 1

?

[hagenw]

No, it would simplify to:

if table.type == 'filewise':
    count += 1

[frankenjoe]

Hehe, of course :)

So yes, I agree this seems like a useful extensions.

[frankenjoe]

Ah wait, now I get it. You want to introduce 'filewise', 'segmented' and 'misc'. I thought you only want to distinguish between Table and MiscTable, i.e. 'table' and 'misc' as type. Mhh, don't we then run again into the issue that MiscTable can have a filewise index :)

[hagenw]

I changed the title of this issue to better track what we want to have.

It's indeed not completely straightforward as before we always derived the table type from the index, compare the current documentation of Table.type:

My proposal would be the following for [MiscTable|Table].type:

• for Table return filewise or segmented
• for MiscTable return misc independent of the used index

I guess that's more complicated from a consistency/documentation point of view, but would made writing code easier.

But I'm also not 100% sure we should add this change.

[frankenjoe]

Yes, I am not sure if it is a good idea to mix up table type and index type in one field. That's why I intuitively expected that type would only have the values table and misc. This would change your example to:

if table.type == 'misc' and table.is_filewise:
    count += 1

[hagenw]

Yes, but Table.type is already in use and returns the index type (which we set identical to table type when the misc table was not existing).

So yes, maybe we don't do anything here and propose to use code like

if isinstance(table, audformat.Table) and table.is_filewise:
    count += 1

[hagenw]

Then you also have the option if you want to count misc tables using a filewise index as well, e.g.

if audformat.is_filewise_index(table.index):
    count += 1

[frankenjoe]

Agree