Re: raw output from copy - Mailing list pgsql-hackers

The regression tests seem to adequately cover all new functionality, though I wonder if we should add some cases that highlight situations where BINARY mode is insufficient.

One thing I tried to test RAW was to load an existing json file.

My own personal test was to load an existing .json file into a 1x1 bytea table, which worked. From there I was able to 

select encode(col_name,'escape')::text::jsonb from test_table

and the json was correctly converted.

A similar test copying binary failed.

A write up of the test looks like this:

\copy (select '{"foo": "bar"}') to '/tmp/raw_test.jsonb' (format raw);

COPY 1

create temporary table raw_byte (b bytea);

CREATE TABLE

create temporary table raw_text (t text);

CREATE TABLE

\copy raw_jsonb from '/tmp/raw_test.blob' (format raw);

psql:/home/ubuntu/raw_test.sql:9: ERROR:  relation "raw_jsonb" does not exist

\copy raw_byte from '/tmp/raw_test.blob' (format raw);

COPY 1

select encode(b,'escape')::text::json from raw_byte;

     encode     

----------------

 {"foo": "bar"}

(1 row)

\copy raw_text from '/tmp/raw_test.blob' (format raw);

COPY 1

select t::jsonb from raw_text;

       t        

----------------

 {"foo": "bar"}

(1 row)

create temporary table binary_byte (b bytea);

CREATE TABLE

create temporary table binary_text (t text);

CREATE TABLE

\copy binary_byte from '/tmp/raw_test.blob' (format binary);

psql:/home/ubuntu/raw_test.sql:22: ERROR:  COPY file signature not recognized

select encode(b,'escape')::jsonb from binary_byte;

 encode 

--------

(0 rows)

\copy binary_text from '/tmp/raw_test.blob' (format binary);

psql:/home/ubuntu/raw_test.sql:26: ERROR:  COPY file signature not recognized

select t::jsonb from binary_text;

 t 

---

(0 rows)

So, if we want to add a regression test to demonstrate to posterity why we need RAW for cases that BINARY can't handle, I offer the attached file.

Does anyone else see value in adding that to the regression tests?

 

Before I give my approval, I want to read it again more closely to make sure that no cases were skipped with regard to the  (binary || raw) and (binary || !raw) tests. Also, I want to use it on some of my problematic files. Maybe I'll find a good edge case. Probably not.

I don't know why I thought this, but when I looked at the patch, I assumed that the ( binary || raw ) tests were part of a large if/elseif/else waterfall. They are not. They stand alone. There are no edge cases to find.

Review complete and passed. I can re-review if we want to add the additional test.