stats: fix panic when dump pseudo columns

[alivxxx]

What problem does this PR solve?

runtime error: invalid memory address or nil pointer dereference
goroutine 25913493 [running]:
net/http.(*conn).serve.func1(0xc30491c000)
	/usr/local/go/src/net/http/server.go:1769 +0xc9
panic(0x1a4f380, 0x2c1a650)
	/usr/local/go/src/runtime/panic.go:522 +0x1b5
github.com/pingcap/tidb/util/chunk.(*Chunk).NumCols(...)
	/home/jenkins/workspace/release_tidb_2.1-ga/go/src/github.com/pingcap/tidb/util/chunk/chunk.go:272
github.com/pingcap/tidb/util/chunk.(*Chunk).NumRows(...)
	/home/jenkins/workspace/release_tidb_2.1-ga/go/src/github.com/pingcap/tidb/util/chunk/chunk.go:277
github.com/pingcap/tidb/util/chunk.(*Iterator4Chunk).Begin(...)
	/home/jenkins/workspace/release_tidb_2.1-ga/go/src/github.com/pingcap/tidb/util/chunk/iterator.go:115
github.com/pingcap/tidb/statistics.(*Histogram).ConvertTo(0xc0076fbb80, 0xc4279c9200, 0xc04faf9c20, 0xc04fca1f50, 0xc4ace6a0c0, 0x0)
	/home/jenkins/workspace/release_tidb_2.1-ga/go/src/github.com/pingcap/tidb/statistics/histogram.go:160 +0xa9
github.com/pingcap/tidb/statistics.(*Handle).tableStatsToJSON(0xc0022b0630, 0xc316905570, 0x6, 0xc004182480, 0xc2e7, 0x160, 0x3, 0x1e247b8)
	/home/jenkins/workspace/release_tidb_2.1-ga/go/src/github.com/pingcap/tidb/statistics/dump.go:103 +0x27e
github.com/pingcap/tidb/statistics.(*Handle).DumpStatsToJSON(0xc0022b0630, 0xc316905570, 0x6, 0xc004182480, 0xc856fe8088, 0xc316905577, 0xb)
	/home/jenkins/workspace/release_tidb_2.1-ga/go/src/github.com/pingcap/tidb/statistics/dump.go:64 +0x33f

What is changed and how it works?

The histograms of the pseudo table are not initialized properly, the bounds of them may be nil, so it will panic when we try to dump the stats.

Check List

Tests

• Unit test

Code changes

• Has exported function/method change

Side effects

• None

Related changes

• None

[alivxxx]

/rebuild

[codecov]

Codecov Report

Merging #11430 into master will not change coverage.
The diff coverage is n/a.

@@             Coverage Diff             @@
##             master     #11430   +/-   ##
===========================================
  Coverage   81.4025%   81.4025%           
===========================================
  Files           424        424           
  Lines         90846      90846           
===========================================
  Hits          73951      73951           
  Misses        11595      11595           
  Partials       5300       5300

[lzmhhh123]

LGTM.

[winoros]

The pseudo estimation with pseudo histograms is different with that without histograms?

[alivxxx]

I think no, this PR only initialize the Bounds to an empty chunk, and there is no code that check if Bounds is nil now.

[eurekaka]

LGTM

[alivxxx]

/run-all-tests

[sre-bot]

/run-all-tests

[sre-bot]

@lamxTyler merge failed.

[alivxxx]

/run-unit-test

[sre-bot]

cherry pick to release-3.0 in PR #11460

[sre-bot]

cherry pick to release-2.1 failed