On Tue, 27 Nov 2007 21:12:00 -0500 Lew <lew@xxxxxxxxxxxxxxxxxxxxxxx> wrote: > Lew wrote: > >> Try eliminating the double quotes in the CSV file. "Wannabe" > >> NULL would then be ,, (consecutive commas) > >> in the CSV. From the docs, you don't even need the "NULL AS" > >> clause in your COPY statement. > > Ivan Sergio Borgonovo wrote: > > Exactly what I did because fortunately there weren't too many > > chances of weird stuff in 2000 records (sed -e 's/,""/,/g'). > > And this worked, right? right and I call it pre-processing. > > I thought it may be missing total support of UTF-8 or if I did > > something wrong or it is actually a "feature". > This clearly has nothing to do with UTF-8, and everything to with > comma-comma representing a NULL and comma-quote-quote-comma > representing an empty string. OK... misinterpreted. I thought that NULL AS '' means ,'',[1] so that empty strings could be imported as NULL if necessary and as empty string if not. So at my understanding there is no way to use \copy and insert NULL when it encounter an empty string and NULL AS '' doesn't do anything in CSV mode since when I have ,, it actually imports NULL and when I have ,'', it imports empty strings that is the same behaviour I get without NULL AS ''. Correct? If it is I found the documentation a bit misleading. I admit it could be due to not being English mother tongue. thx [1] I did try with "''", '""', '\"\"'... -- Ivan Sergio Borgonovo http://www.webthatworks.it ---------------------------(end of broadcast)--------------------------- TIP 3: Have you checked our extensive FAQ? http://www.postgresql.org/docs/faq
