Control dtype with ascii.read (converters API needs better doc or refactoring)

I cannot find a way to control the dtype of the output table when reading a file into a Table. Consider the following MWE, with 3 numerical columns, while one of them would preferably be kept as a string:

>>> from astropy.io import ascii

>>> indata = ("# This is a dummy file\n" 
...           "# with some text to ignore, and a header with column names\n" 
...           "# ra dec objid\n" 
...           "1 2 345\n" 
...           "3 4 456\n") 

>>> ascii.read(indata, format='commented_header', header_start=2, guess=False, fast_reader=False)
<Table length=2>
  ra   dec  objid
int64 int64 int64
----- ----- -----
    1     2   345
    3     4   456

>>> ascii.read(indata, format='commented_header', header_start=2, dtye=('i8', 'i8', 'S10'), guess=False, fast_reader=False)
TypeError: __init__() got an unexpected keyword argument 'dtye'

Reading in the same with np.loadtxt and then converting to a Table works, but it should ideally be supported directly.

import numpy as np
from astropy.table import Table

>>> Table(np.loadtxt('/tmp/a', dtype=[('ra', 'i8'), ('dec', 'i8'), ('objid', 'S10')]))
<Table length=2>
  ra   dec   objid 
int64 int64 bytes10
----- ----- -------
    1     2     345
    3     4     456

[taldcroft]

I’ve implemented this in a few lines of code. As always the pain is testing, docs etc. But maybe there will be a PR on the way.

In [2]: >>> ascii.read(indata, format='commented_header', header_start=2, dtype=('i8', 'i4', 'S10'), guess=False, fast_reader=False)
Out[2]: 
<Table length=2>
  ra   dec   objid 
int64 int32 bytes10
----- ----- -------
    1     2     345
    3     4     456

[bsipocz]

Oh, yes, indeed, this is exactly what I need. One minor comment though, it would be helpful to have the word dtype somewhere in the docs, as I was searching for dtype in that and many other docs pages without any useful results. (maybe it’s just me, that case this can be closed without a docs change, otherwise this can be a good “first issue”).

It’s also not clear what the “previous section” is referred to in These take advantage of the convert_numpy() function which returns a two-element tuple (converter_func, converter_type) as described in the previous section.