Accepted filenames rules

[pskoda]

Hi Mike,

After some searching in forums I do not still have complete answer for the following question so I have to ask here.

We are now considering radical changes in our CCD acquisition system and people want to introduce new more verbose filenames instead of DOS compatible ab120015.fit where a=1990 b=February for 15th file in night 12nd Feb 1990 introduced many years ago.

We have various weird suggestions like "2012-02.25_0012.fit" or "a201202250012.fit" or "2012-02-25_0012.fit" etc .....

I suppose some of that filenames formats might introduce hidden errors in processing of such filenames in IRAF or in storing them in databases and using as access reference in VO tables (namely SSA accref) and so forth.

I remember there were even troubles in IRAF with capital letters in names or extensions.

Could you please, summarize both the rules IRAF accepts (and if there were more restriction in older versions of IRAF) and your recommendation what to avoid with respect to the VO capabilities (and even planned SSA support in future IRAF) ?

Best regards

Petr Skoda

[fitz]

Hi Petr,

Strictly speaking, (practically) anything that is a legal filename for the host platform is legal as an IRAF filename. So, upper/lowercase letters, numbers, special chars like '.', '-', or '_' can be used to form a valid name.

That said, the recommendation has traditionally been to begin an image name with a letter to avoid confusion. The reason is that image name can be referred to without the explicit extension, and so a name like "1234.fits" can be passed into a task as simply "1234", and depending on the task the CL may confuse this with an integer value 1234. To avoid the ambiguity, you can quote the "1234", but that gets painful and it is just easier to avoid. For the same reasons, tasks that parse text files like solution databases may similarly get confused, and extremely long names are unpleasant for users.

Use of capitals in image extensions is on the list of improvements but hasn't been implemented yet. Data access for SSA/SIA services is handled separately and the rules for URL access (e.g. to allow a '?' or '&' in the URL) are implemented in the CL directly and not the system file handling.

What I'd suggest if you're contemplating a new naming scheme is to generate some sample files and see what breaks. Certainly avoid special chars like quotes and semi-colons, there is also no support currently in the system for spaces in file/path names.

Cheers,
-Mike

[pskoda]

Thank you, Mike
I will later put here my observations if something unexpected happens to warn others.

Best regards

Petr