sql: various COPY improvements

[otan]

refs #41608

Release note: None

[cockroach-teamcity]

This change is 

[rafiss]

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @otan and @rafiss)

---

pkg/sql/copy.go line 160 at r2 (raw file):

		if c.format == tree.CopyFormatBinary {
			return nil, pgerror.Newf(
				pgcode.FeatureNotSupported,

PG says "This option is not allowed when using binary format." does PG return FeatureNotSupported here too? i'm thinking InvalidParameter may make more sense

---

pkg/sql/copy.go line 174 at r2 (raw file):

		if len(delim) != 1 || !utf8.ValidString(delim) {
			return nil, pgerror.Newf(
				pgcode.FeatureNotSupported,

i think PG also doesn't allow this, so ditto

---

pkg/sql/copy.go line 183 at r2 (raw file):

		if c.format == tree.CopyFormatBinary {
			return nil, pgerror.Newf(
				pgcode.FeatureNotSupported,

ditto

[otan]

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @rafiss)

---

pkg/sql/copy.go line 160 at r2 (raw file):

otan=# \set VERBOSITY verbose
otan=# copy t from stdin delimiter ',';
Enter data to be copied followed by a newline.
End with a backslash and a period on a line by itself, or an EOF signal.
>> COPY 0

totally an us problem

---

pkg/sql/copy.go line 174 at r2 (raw file):

Previously, rafiss (Rafi Shamim) wrote…

i think PG also doesn't allow this, so ditto

otan=# copy t from stdin delimiter 'xx';
ERROR:  0A000: COPY delimiter must be a single one-byte character
LOCATION:  ProcessCopyOptions, copy.c:561

this is correct

---

pkg/sql/copy.go line 183 at r2 (raw file):

otan=# copy t from stdin null 'xxx';
Enter data to be copied followed by a newline.
End with a backslash and a period on a line by itself, or an EOF signal.
>> COPY 0

totally an us problem

[rafiss]

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @otan and @rafiss)

---

pkg/sql/copy.go line 160 at r2 (raw file):

Previously, otan (Oliver Tan) wrote…

otan=# \set VERBOSITY verbose
otan=# copy t from stdin delimiter ',';
Enter data to be copied followed by a newline.
End with a backslash and a period on a line by itself, or an EOF signal.
>> COPY 0

totally an us problem

but you'd need to test with copy t from stdin binary delimiter ',';

---

pkg/sql/copy.go line 183 at r2 (raw file):

Previously, otan (Oliver Tan) wrote…

otan=# copy t from stdin null 'xxx';
Enter data to be copied followed by a newline.
End with a backslash and a period on a line by itself, or an EOF signal.
>> COPY 0

totally an us problem

but what does PG say when you try with binary format?

[rafiss]

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @otan and @rafiss)

---

pkg/sql/copy.go line 160 at r2 (raw file):

Previously, rafiss (Rafi Shamim) wrote…

but you'd need to test with copy t from stdin binary delimiter ',';

just checked -- it's 42601 syntax_error

---

pkg/sql/copy.go line 174 at r2 (raw file):

Previously, otan (Oliver Tan) wrote…

otan=# copy t from stdin delimiter 'xx';
ERROR:  0A000: COPY delimiter must be a single one-byte character
LOCATION:  ProcessCopyOptions, copy.c:561

this is correct

you're right PG does use 0A000 for this one

---

pkg/sql/copy.go line 183 at r2 (raw file):

Previously, rafiss (Rafi Shamim) wrote…

but what does PG say when you try with binary format?

this is also it's 42601 syntax_error

[otan]

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @rafiss)

---

pkg/sql/copy.go line 160 at r2 (raw file):

Previously, rafiss (Rafi Shamim) wrote…

just checked -- it's 42601 syntax_error

Done.

---

pkg/sql/copy.go line 183 at r2 (raw file):

Previously, rafiss (Rafi Shamim) wrote…

this is also it's 42601 syntax_error

ah they must do this at the parser layer. ok

[otan]

bors r=rafiss

[craig]

Build failed (retrying...):

• GitHub CI (Cockroach)

[craig]

Build succeeded:

• GitHub CI (Cockroach)