ASCII Table tricks

[Thomas A. McGlynn]

RANDY THOMPSON (301) 286-8800 wrote:
>... However,
> it seems to me that a Binary Tables extension could be used to store the
> data described earlier by Steve Allen? Maybe I'm missing something (since I
> haven't seen this suggested already), but why not store the ASCII text as
> character strings and define fields of byte data type to contain the byte
> values of non-printable ASCII characters? If you strip off the headers,
> the remaining records should be identical to those that could be produced
> with the ASCII Tables extension.
>    If this isn't suitable, I would think one could still propose a change to
> the current FITS standard that would (unambiguously) allow non-printable ASCII
> characters as non-field values. If existing FITS readers are really not
> impacted, why should such a proposed change not be approved? Of course, the
> change would still need to go through the established approval process.
...
> 
> Randy Thompson

I think that adding additional meaningless columns to a binary
table does substantial violence to the file since, e.g., it changes
the column numbers of the real data.  But I really like the idea that
ASCII tables be readable as readable as possible.  When ASCII tables
were introduced they were the only standard way for creating tabular
data.  But with the
advent of binary tables the only rationale I can see for using them
is that they are directly readable by humans.  I can attest from
personal
experience that it is very helpful to be able to read the tables -- one
trick I use is to adjust the width of my screen to the length of the
table.
Even if the rows don't start in the beginning it still helps a lot.

In writing a STBRYAFWFIDL (Someday-to-be-released,
Yet-another-FITS-writer-
for-IDL), I had anticipated users wanting to separate and delimit rows
and I'm quite disappointed that it is at best questionable FITS to
delimit the columns in the most natural way.

I think we are choosing between making ASCII tables completely obsolete
or leaving them a niche market in the FITS file economy.  If it truly
is 'illegal' to use \n in a FITS table, I would second Randy's call
for an amendment to their definition.

		Tom McGlynn
		GSFC.

		Tom McGlynn