If I put show[piece_data.tmini] directly after the extract expression, it prints:
("%%IMG2ADAAMAD/AP8A/wBHAQEALgEDAC0BAQAvAQEAAQEBAC0BAQACAQEAKwEBAAYBAgAmAQMAAgEBAAIBAgAmAQEAAQEBAAMBAgABAQEAJgEBAAEBAQADAQIAAgEBACcBAQAEAQEAAgEBACcBAQAIAQEALwEBAC8BAQAwAQEA/wD/AP8ARQ==")
…which represents a list containing a single string (even though that’s not the syntax for creating a list containing a single string). Let’s call it imglist:
imglist:list "%%IMG2ADAAMAD/AP8A/wBHAQEALgEDAC0BAQAvAQEAAQEBAC0BAQACAQEAKwEBAAYBAgAmAQMAAgEBAAIBAgAmAQEAAQEBAAMBAgABAQEAJgEBAAEBAQADAQIAAgEBACcBAQAEAQEAAgEBACcBAQAIAQEALwEBAC8BAQAwAQEA/wD/AP8ARQ=="
Now print[imglist] will try to convert the list to a string, which works and produces the intended result (as you saw with your “this is what I want” message). However, image[imglist] wants its argument to be an x,y size, and only falls back to trying to interpret a string if it can’t do that. The argument it gets is a list, just not a list of two integers, so you wind up with an 0x0 image.
In general, you probably want to change the way you extract the row from the table:
piece_data: first rows select
tname:tname
tdesc:tdesc
tmini:tmini
where index=row
from me.value
Now piece_data is a dictionary, as you expected, rather than a table.
In this specific instance, since this is the click handler of a grid, we can assume that the relevant row of the grid is already selected, and we can just say:
piece_data:me.rowvalue
And now the rest of your code should work as you expect.