Re: COPY CSV keywords

Bruce Momjian wrote:

>[ discussion moved to hackers.]
>
>Peter Eisentraut wrote:
>
>
>>Bruce Momjian wrote:
>>
>>
>>>I have applied the attached patch that complete TODO item:
>>>
>>> o -Allow dump/load of CSV format
>>>
>>>This adds new keywords to COPY and \copy:
>>>
>>> CSV - enable CSV mode
>>> QUOTE - specify quote character
>>> ESCAPE - specify escape character
>>> FORCE - force quoting of specified columns
>>>
>>>
>>FORCE QUOTE
>>
>>
>>
>>> LITERAL - prevent NULL checks for specific columns
>>>
>>>
>>NO NULL CHECK
>>
>>
>
>I considered going this direction, but it broke the WITH clause style of
>COPY. Previously it was "WITH keyword value". Now it is also "WITH
>keyword value, value" too. This would add "WITH keyword keyword value,
>value".
>
>It would change:
>
> COPY pg_language TO '/tmp/x' WITH CSV FORCE lanname
> LITERAL lanacl
>to:
> COPY pg_language TO '/tmp/x' WITH CSV FORCE QUOTE lanname NO
> NULL CHECK lanacl
>
>If folks like it, I can make the change.
>
>
>
[reposting my comments from patches]

The problem is that using QUOTE or NULL in these phrases might look
confusing, e.g.
COPY mytable TO 'mytable.csv' CSV QUOTE '"' FORCE QUOTE field1,field2;
COPY mytable FROM 'mytable.csv' NULL '' CSV NO NULL CHECK field1,field2;

I also don't think NO NULL CHECK actually matches the use case for this
very well (and I'm dubious about LITERAL too). They both describe the
actual behaviour, but not what you are trying to achieve. Essentially,
this would be used when you have a field with a NOT NULL constraint, but
the input CSV data stream has what would otherwise be considered nulls.
(COPY itself will never produce such a CSV, as non-null values that
resemble null are always quoted, but third party programs well might.)
So an alternative might be FORCE NOT NULL, but for the previous
consideration. Perhaps use of an optional preposition might make things
slightly clearer, e.g.:

COPY mytable TO 'mytable.csv' CSV QUOTE '"' FORCE QUOTE IN
field1,field2;
COPY mytable FROM 'mytable.csv' NULL '' CSV FORCE NOT NULL IN
field1,field2;

But it does start to look a little too much like COBOL .

So I'm interested to see if there are any other inspirations people have.

cheers

andrew