On 14/01/11 12:42, Pádraig Brady wrote: > On 12/01/11 18:04, Assaf Gordon wrote: >> Pádraig Brady wrote, On 01/11/2011 07:35 AM: >>> >>> Spending another few minutes on this, I realized >>> that we should not be trying to homogenize the number >>> of fields from each file, but rather the fields used >>> for a particular file in each line. The only sensible >>> basis for that is the first line as previously suggested. >>> >>> The interface would be a little different for that. >>> I was thinking of: >>> >>> -o 'header' Infer the format from the first line of each file >>> >> I second the idea of using the first line as the basis for the >> auto-formatting, >> but have reservation about the wording: '-o header' somewhat implies that >> the first line has to be an actual header line (with column names or >> similar), while it can just be the first line of actual data if the file >> doesn't have a header line. >> >> Something like '-o auto' might be less confusing. > > I've attached a patch that implements the above. > I'll hold off applying for a short while as > we might be doing a bug fix release.
Actually Jim said this feature is Ok to add now. I also need to amend NEWS to describe it as a "New Feature" rather than as a "Change in behavior". cheers, Pádraig.
