I wrote:
> Peter Eisentraut <peter_e@gmx.net> writes:
>> Maybe wdiff will do what you want.
> I don't think we'd consider making the kind of changes to the COPY
> specification that would be needed to allow this sort of thing in COPY
> mode. However, it'd be pretty trivial to put \n instead of space
> between fields in the INSERT dump formats. I wonder whether that would
> answer the need ...
On further reflection I think Peter's suggestion is better. It's not
hard to think of cases where a data-value-per-line format is *less*
useful not more so for diff-ing, because you lose all sense that a
table row is one logical unit. What the OP seems to be concerned about
is insertion/modification/deletion of words within fairly long data
fields. wdiff is ideal for that problem, while simply breaking at field
boundaries wouldn't help all that much.
regards, tom lane