sql: enable manual control of column families

[danhhz]

• Switch to the new FormatVersion
• Hook up the sql syntax in CREATE TABLE
• Update SHOW CREATE TABLE
• Update the WideTable benchmark to take advantage of column families

name old time/op new time/op delta
WideTable1_Cockroach-8 1.88ms ± 1% 1.50ms ± 0% -20.08% (p=0.008 n=5+5)
WideTable10_Cockroach-8 5.86ms ± 1% 2.59ms ± 1% -55.77% (p=0.016 n=4+5)
WideTable100_Cockroach-8 49.1ms ± 5% 13.2ms ± 5% -73.07% (p=0.008 n=5+5)
WideTable1000_Cockroach-8 570ms ± 5% 120ms ± 3% -79.00% (p=0.008 n=5+5)

name old alloc/op new alloc/op delta
WideTable1_Cockroach-8 265kB ± 0% 163kB ± 0% -38.55% (p=0.008 n=5+5)
WideTable10_Cockroach-8 1.43MB ± 0% 0.37MB ± 0% -74.27% (p=0.008 n=5+5)
WideTable100_Cockroach-8 13.2MB ± 0% 2.4MB ± 0% -82.00% (p=0.016 n=4+5)
WideTable1000_Cockroach-8 175MB ± 0% 21MB ± 0% -87.76% (p=0.008 n=5+5)

name old allocs/op new allocs/op delta
WideTable1_Cockroach-8 2.19k ± 0% 1.91k ± 0% -12.88% (p=0.008 n=5+5)
WideTable10_Cockroach-8 6.59k ± 0% 3.96k ± 0% -39.83% (p=0.008 n=5+5)
WideTable100_Cockroach-8 50.2k ± 0% 23.9k ± 0% -52.40% (p=0.008 n=5+5)
WideTable1000_Cockroach-8 546k ± 0% 223k ± 0% -59.09% (p=0.008 n=5+5)

---

This change is

[petermattis]

4.75x faster on the WideTable1000 benchmark. Not bad. Not bad at all.

---

Review status: 0 of 14 files reviewed at latest revision, all discussions resolved, some commit checks failed.

---

Comments from Reviewable

[RaduBerinde]

Awesome!

---

Review status: 0 of 14 files reviewed at latest revision, 1 unresolved discussion, some commit checks failed.

---

sql/show.go, line 155 [r1] (raw file):

      )
  }
  for _, fam := range desc.Families {

Should we omit single-column families from the output? (Creating a table with and without explicitly specifying single-column families should be pretty much the same thing right?)

---

Comments from Reviewable

[petermattis]

Review status: 0 of 14 files reviewed at latest revision, 2 unresolved discussions, some commit checks failed.

---

sql/testdata/family, line 10 [r1] (raw file):

  d INT,
  FAMILY f1 (a, b),
  FAMILY f2 (c, d)

Do you have any test cases for the automatically generated family names? That is, for something like FAMILY (c, d)?

---

Comments from Reviewable

[danhhz]

Review status: 0 of 15 files reviewed at latest revision, 2 unresolved discussions, some commit checks pending.

---

sql/show.go, line 155 [r1] (raw file):

Previously, RaduBerinde wrote…

Should we omit single-column families from the output? (Creating a table with and without explicitly specifying single-column families should be pretty much the same thing right?)

If a table has two single column families and you use the output of SHOW CREATE TABLE to make it again in a later version of cockroach, they may be placed in the same family by the upcoming heuristics. I'd rather err on the side of making this as roundtrippable as possible.

---

sql/testdata/family, line 10 [r1] (raw file):

Previously, petermattis (Peter Mattis) wrote…

Do you have any test cases for the automatically generated family names? That is, for something like FAMILY (c, d)?

Not end to end. Done.

---

Comments from Reviewable

[spencerkimball]

👍 Nice!!!