sql: internal error: (-0E-2019):::DECIMAL: could not evaluate -0E-2019 as Datum type DDecimal from string "0E-2019": could not parse "0E-2019" as type decimal #40929

maddyblue · 2019-09-19T20:57:24Z

CREATE TABLE IF NOT EXISTS tab_orig AS
	SELECT g FROM generate_series(NULL, NULL) AS g;

SELECT 0 / '-Infinity'::DECIMAL FROM tab_orig;

stack trace:
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/execinfra/expr.go:78: processExpression()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/execinfra/expr.go:161: Init()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/execinfra/processorsbase.go:145: Init()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/execinfra/processorsbase.go:776: InitWithEvalCtx()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/execinfra/processorsbase.go:752: Init()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/rowexec/tablereader.go:92: newTableReader()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/rowexec/processors.go:138: NewProcessor()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/rowflow/row_based_flow.go:165: makeProcessor()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/rowflow/row_based_flow.go:65: setupProcessors()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/rowflow/row_based_flow.go:48: Setup()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/distsql/server.go:331: setupFlow()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/distsql/server.go:404: SetupLocalSyncFlow()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/distsql_running.go:241: setupFlows()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/distsql_running.go:330: Run()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/distsql_running.go:966: PlanAndRun()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/conn_executor_exec.go:826: execWithDistSQLEngine()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/conn_executor_exec.go:718: dispatchToExecutionEngine()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/conn_executor_exec.go:417: execStmtInOpenState()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/conn_executor_exec.go:98: execStmt()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/conn_executor.go:1238: execCmd()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/conn_executor.go:1167: run()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/conn_executor.go:444: ServeConn()
/home/mjibson/go/src/github.com/cockroachdb/cockroach/pkg/sql/pgwire/conn.go:584: func1()
runtime/asm_amd64.s:1337: goexit()

Jira issue: CRDB-5499

The text was updated successfully, but these errors were encountered:

rafiss · 2019-09-23T18:29:46Z

This started happening because of the change made in #40327

That PR changes func (d *DDecimal) SetString(s string) error in datum.go so that it uses ExactCtx instead of HighPrecisionCtx.

I think the issue is that the division is performed using DecimalCtx, which results in a value of 0E-2019, but ExactCtx is not able to handle that value.

maddyblue · 2019-10-26T00:44:04Z

The postgres spec for decimals is "up to 131072 digits before the decimal point; up to 16383 digits after the decimal point" (used when there is no user-specified precision) and "the maximum allowed precision when explicitly specified in the type declaration is 1000".

We could constrain our decimal use to those specs, and in addition remove our support for infinity, which is not supported by postgres decimals (and thus has no binary encoding).

maddyblue · 2019-10-26T00:45:01Z

As a side effect of not having infinity, the above query will stop failing for this reason.

This commit skips failing `sqlsmith` roachtest because of a known issue (cockroachdb#40929) in order to avoid unnecessary noise. Release note: None

This commit makes `sqlsmith` roachtest not fail on one known issue (cockroachdb#40929) in order to avoid unnecessary noise. Release note: None

otan · 2023-05-22T05:17:39Z

this no longer panics, but produces an incorrect result. marking as duplicate of #103633

yuzefovich · 2023-05-30T07:45:38Z

It's still reproducible in distributed setting (as Rafi pointed out above).

Previously, the logic for decimal and float division, floor division and mod operators was incorrect for a few edge cases involving `NaN` or `Infinity` values. For example, `'NaN'::DECIMAL / 0` would throw a division-by-zero error when it should evaluate to `NaN` and `0/'inf'::DECIMAL` returned `0E-2019` instead of just `0`. This patch updates the special-case logic to mirror that of postgres, so division-by-zero errors always check the `NaN` case and the division by infinity case returns a `0` without extra digits. Fixes cockroachdb#40929 Fixes cockroachdb#103633 Release note (bug fix): Fixed edge cases in decimal and float evaluation for division operators. `'NaN'::DECIMAL / 0` will now return `NaN` instead of a division-by-zero error, and `0 / 'inf'::DECIMAL` will return `0` instead of `0E-2019`.

106472: sql: fix decimal evaluation edge cases r=DrewKimball a=DrewKimball Previously, the logic for decimal and float division, floor division and mod operators was incorrect for a few edge cases involving `NaN` or `Infinity` values. For example, `'NaN'::DECIMAL / 0` would throw a division-by-zero error when it should evaluate to `NaN` and `0/'inf'::DECIMAL` returned `0E-2019` instead of just `0`. This patch updates the special-case logic to mirror that of postgres, so division-by-zero errors always check the `NaN` case and the division by infinity case returns a `0` without extra digits. Fixes #40929 Fixes #103633 Release note (bug fix): Fixed edge cases in decimal and float evaluation for division operators. `'NaN'::DECIMAL / 0` will now return `NaN` instead of a division-by-zero error, and `0 / 'inf'::DECIMAL` will return `0` instead of `0E-2019`. 106790: tree: fix statement tag for tree.ShowSequences r=knz,rafiss a=andyyang890 Epic: None Release note (bug fix): The statement tag for `SHOW SEQUENCES` is now corrected to be `SHOW SEQUENCES` instead of `SHOW SCHEMAS`. Co-authored-by: Drew Kimball <[email protected]> Co-authored-by: Andy Yang <[email protected]>

Previously, the logic for decimal and float division, floor division and mod operators was incorrect for a few edge cases involving `NaN` or `Infinity` values. For example, `'NaN'::DECIMAL / 0` would throw a division-by-zero error when it should evaluate to `NaN` and `0/'inf'::DECIMAL` returned `0E-2019` instead of just `0`. This patch updates the special-case logic to mirror that of postgres, so division-by-zero errors always check the `NaN` case and the division by infinity case returns a `0` without extra digits. Fixes cockroachdb#40929 Fixes cockroachdb#103633 Release note (bug fix): Fixed edge cases in decimal and float evaluation for division operators. `'NaN'::DECIMAL / 0` will now return `NaN` instead of a division-by-zero error, and `0 / 'inf'::DECIMAL` will return `0` instead of `0E-2019`.

maddyblue added C-bug Code not up to spec/doc, specs & docs deemed correct. Solution expected to change code/behavior. O-sqlsmith labels Sep 19, 2019

maddyblue assigned rafiss Sep 19, 2019

rafiss added the S-3-erroneous-edge-case Database produces or stores erroneous data without visible error/warning, in rare edge cases. label Sep 24, 2019

rafiss removed their assignment Jan 28, 2020

yuzefovich mentioned this issue Apr 23, 2020

roachtest: sqlsmith/setup=seed/setting=no-mutations failed #47397

Closed

asubiotto mentioned this issue Sep 1, 2020

roachtest: sqlsmith/setup=rand-tables/setting=no-mutations failed #53761

Closed

yuzefovich mentioned this issue Sep 24, 2020

roachtest: could not parse "0E-2019" as type decimal #54731

Closed

asubiotto mentioned this issue Oct 6, 2020

roachtest: sqlsmith/setup=rand-tables/setting=no-ddl failed #55236

Closed

yuzefovich mentioned this issue Oct 10, 2020

roachtest: sqlsmith/setup=seed/setting=no-mutations failed #55405

Closed

This was referenced Nov 10, 2020

roachtest: sqlsmith/setup=rand-tables/setting=no-ddl failed #56485

Closed

roachtest: sqlsmith/setup=rand-tables/setting=no-ddl failed #56641

Closed

asubiotto mentioned this issue Nov 30, 2020

roachtest: sqlsmith/setup=seed/setting=no-mutations failed #57209

Closed

asubiotto mentioned this issue Mar 1, 2021

roachtest: sqlsmith/setup=seed/setting=no-mutations failed #61228

Closed

yuzefovich mentioned this issue Mar 14, 2021

roachtest: sqlsmith/setup=seed/setting=no-ddl failed #61978

Closed

yuzefovich mentioned this issue Mar 22, 2021

roachtest: sqlsmith/setup=rand-tables/setting=no-mutations failed #62292

Closed

yuzefovich added a commit to yuzefovich/cockroach that referenced this issue Mar 22, 2021

roachtest: skip expected error in sqlsmith

8aa000d

This commit skips failing `sqlsmith` roachtest because of a known issue (cockroachdb#40929) in order to avoid unnecessary noise. Release note: None

yuzefovich mentioned this issue Mar 22, 2021

roachtest: not fail sqlsmith on known issue #62402

Merged

yuzefovich mentioned this issue Mar 22, 2021

release-20.2: roachtest: improve sqlsmith #62403

Merged

yuzefovich added a commit to yuzefovich/cockroach that referenced this issue Mar 23, 2021

roachtest: not fail sqlsmith on known issue

e6c5332

This commit makes `sqlsmith` roachtest not fail on one known issue (cockroachdb#40929) in order to avoid unnecessary noise. Release note: None

yuzefovich mentioned this issue May 19, 2023

sql: 0::decimal/infinity returns 0E-2019 instead of 0 #103633

Closed

otan closed this as not planned Won't fix, can't repro, duplicate, stale May 22, 2023

exalate-issue-sync bot closed this as completed May 22, 2023

yuzefovich reopened this May 30, 2023

yuzefovich mentioned this issue May 30, 2023

sqlsmith: unable to vectorize execution plan: (-0.0...000000" as type decimal #103820

Closed

yuzefovich mentioned this issue Jul 5, 2023

roachtest: sqlsmith/setup=rand-tables/setting=no-ddl failed #106048

Closed

yuzefovich added the C-test-failure Broken test (automatically or manually discovered). label Jul 6, 2023

DrewKimball mentioned this issue Jul 8, 2023

sql: fix decimal evaluation edge cases #106472

Merged

DrewKimball self-assigned this Jul 11, 2023

craig bot closed this as completed in a8b42c9 Jul 14, 2023

mgartner added this to SQL Queries Jul 24, 2023

mgartner moved this to Done in SQL Queries Jul 24, 2023

This was referenced Sep 9, 2023

release-23.1: sql: fix decimal evaluation edge cases #110296

Merged

release-22.2: sql: fix decimal evaluation edge cases #110298

Merged

yuzefovich mentioned this issue May 22, 2024

roachtest: sqlsmith/setup=rand-tables/setting=no-ddl failed #124548

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

sql: internal error: (-0E-2019):::DECIMAL: could not evaluate -0E-2019 as Datum type DDecimal from string "0E-2019": could not parse "0E-2019" as type decimal #40929

sql: internal error: (-0E-2019):::DECIMAL: could not evaluate -0E-2019 as Datum type DDecimal from string "0E-2019": could not parse "0E-2019" as type decimal #40929

maddyblue commented Sep 19, 2019 •

edited by cockroach-jira-scripts

Loading

rafiss commented Sep 23, 2019

maddyblue commented Oct 26, 2019

maddyblue commented Oct 26, 2019

otan commented May 22, 2023

yuzefovich commented May 30, 2023

sql: internal error: (-0E-2019):::DECIMAL: could not evaluate -0E-2019 as Datum type DDecimal from string "0E-2019": could not parse "0E-2019" as type decimal #40929

sql: internal error: (-0E-2019):::DECIMAL: could not evaluate -0E-2019 as Datum type DDecimal from string "0E-2019": could not parse "0E-2019" as type decimal #40929

Comments

maddyblue commented Sep 19, 2019 • edited by cockroach-jira-scripts Loading

rafiss commented Sep 23, 2019

maddyblue commented Oct 26, 2019

maddyblue commented Oct 26, 2019

otan commented May 22, 2023

yuzefovich commented May 30, 2023

maddyblue commented Sep 19, 2019 •

edited by cockroach-jira-scripts

Loading