Bug with CountTable []= 0

[Titousensei]

I think I found a bug with CountTable. In some conditions, when changing the count of one item to zero, the table gets corrupted and another item is set to zero when doing a table lookup (but not from iterator).

Here's a small repro:

import tables

var t = initCountTable[string]()
t[" a s </w> "] = 40335062
t[" t h e r e </w> "] = 23199253
t[" p l a y </w> "] = 7258802
echo t.len

t[" t h e r e </w> "] = 0
echo t.len
echo t

for word, freq in t:
    echo word, "|", freq, " ", t[word]

prints the following:

3
3
{" a s </w> ": 40335062, " p l a y </w> ": 7258802}
 a s </w> |40335062 40335062
 p l a y </w> |7258802 0

As you can see, freq from the iterator is correct, but t[word] is zero.

Also, the len did not change and is inconsistent with the printed table. ( existing issue #12813 )

This is with nim 1.0.4. I can see that upcoming 1.0.6 will have del implemented to replace []= 0, and that might fix this bug. (Or maybe not.)

[Araq]

In my world "counting" never resets values back to 0. Better use a real Table instead.

[Titousensei]

@Araq isn't it bad to have a seemingly normal operator silently corrupt the table? If this is not allowed, the = operator should fail if the value already exists.

[Araq]

Oh I completely agree it's a bug and not the first one, apparently our CountTable stinks ... when you don't use it for counting. :-)