Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[SPARK-22036][SQL] Decimal multiplication with high precision/scale often returns NULL #20023

Closed
wants to merge 17 commits into from
Closed
Show file tree
Hide file tree
Changes from 4 commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,7 @@ import org.apache.spark.sql.catalyst.expressions._
import org.apache.spark.sql.catalyst.expressions.Literal._
import org.apache.spark.sql.catalyst.plans.logical.LogicalPlan
import org.apache.spark.sql.catalyst.rules.Rule
import org.apache.spark.sql.internal.SQLConf
import org.apache.spark.sql.types._


Expand All @@ -42,8 +43,10 @@ import org.apache.spark.sql.types._
* e1 / e2 p1 - s1 + s2 + max(6, s1 + p2 + 1) max(6, s1 + p2 + 1)
* e1 % e2 min(p1-s1, p2-s2) + max(s1, s2) max(s1, s2)
* e1 union e2 max(s1, s2) + max(p1-s1, p2-s2) max(s1, s2)
* sum(e1) p1 + 10 s1
* avg(e1) p1 + 4 s1 + 4
*
* When `spark.sql.decimalOperations.allowTruncat` is set to true, if the precision / scale needed
* are out of the range of available values, the scale is reduced up to 6, in order to prevent the
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

it's better to push some reference about where we get this rule.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

if scale needs to be less than 6, we would fail the analysis, right?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

no, we would return the a type with scale 6 and in CheckOverflow all the numbers which don't fit will be translated to NULL, since this is current Spark behavior in such cases.

May I kindly ask you to elaborate a but more what you mean by "push some reference"? Thanks.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

sorry, typo, "put some reference"...

We can put a link to a document of a mainstream RDBMS and say this rule follows xxx...

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I did, but @gatorsmile told me to remove it: #20023 (comment).

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Did any open source RDBMS have this rule?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Actually we already referred a commercial RDBMS in L33...

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hive has it, but in the documentation it is not explained. And in the comments it just references the blog I referenced, but then removed according to @gatorsmile's comment.

* truncation of the integer part of the decimals.
*
* To implement the rules for fixed-precision types, we introduce casts to turn them to unlimited
* precision, do the math on unlimited-precision numbers, then introduce casts back to the
Expand All @@ -56,6 +59,7 @@ import org.apache.spark.sql.types._
* - INT gets turned into DECIMAL(10, 0)
* - LONG gets turned into DECIMAL(20, 0)
* - FLOAT and DOUBLE cause fixed-length decimals to turn into DOUBLE
* - Literals INT and LONG gets turned into DECIMAL with the precision strictly needed by the value
*/
// scalastyle:on
object DecimalPrecision extends TypeCoercionRule {
Expand Down Expand Up @@ -93,41 +97,76 @@ object DecimalPrecision extends TypeCoercionRule {
case e: BinaryArithmetic if e.left.isInstanceOf[PromotePrecision] => e

case Add(e1 @ DecimalType.Expression(p1, s1), e2 @ DecimalType.Expression(p2, s2)) =>
val dt = DecimalType.bounded(max(s1, s2) + max(p1 - s1, p2 - s2) + 1, max(s1, s2))
CheckOverflow(Add(promotePrecision(e1, dt), promotePrecision(e2, dt)), dt)
val resultType = if (SQLConf.get.decimalOperationsAllowTruncat) {
val resultScale = max(s1, s2)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

we can put this before the if

DecimalType.adjustPrecisionScale(max(p1 - s1, p2 - s2) + resultScale + 1,
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is an example. adjustPrecisionScale is also be applied for all the operations. However, the doc shows the adjustment is only applicable to multiplication and division.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

yes, that may be a difference indeed. But I think it is a minor one, since 99% of the cases the precision is exceeded only in multiplications and divisions.

Copy link
Member

@gatorsmile gatorsmile Jan 16, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We need to make a decision. You know, we try our best to keep our rule as stable as possible.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'm ok to do the adjustment for all operations, which is same as Hive.

resultScale)
} else {
DecimalType.bounded(max(s1, s2) + max(p1 - s1, p2 - s2) + 1, max(s1, s2))
}
CheckOverflow(Add(promotePrecision(e1, resultType), promotePrecision(e2, resultType)),
resultType)

case Subtract(e1 @ DecimalType.Expression(p1, s1), e2 @ DecimalType.Expression(p2, s2)) =>
val dt = DecimalType.bounded(max(s1, s2) + max(p1 - s1, p2 - s2) + 1, max(s1, s2))
CheckOverflow(Subtract(promotePrecision(e1, dt), promotePrecision(e2, dt)), dt)
val resultType = if (SQLConf.get.decimalOperationsAllowTruncat) {
val resultScale = max(s1, s2)
DecimalType.adjustPrecisionScale(max(p1 - s1, p2 - s2) + resultScale + 1,
resultScale)
} else {
DecimalType.bounded(max(s1, s2) + max(p1 - s1, p2 - s2) + 1, max(s1, s2))
}
CheckOverflow(Subtract(promotePrecision(e1, resultType), promotePrecision(e2, resultType)),
resultType)

case Multiply(e1 @ DecimalType.Expression(p1, s1), e2 @ DecimalType.Expression(p2, s2)) =>
val resultType = DecimalType.bounded(p1 + p2 + 1, s1 + s2)
val resultType = if (SQLConf.get.decimalOperationsAllowTruncat) {
DecimalType.adjustPrecisionScale(p1 + p2 + 1, s1 + s2)
} else {
DecimalType.bounded(p1 + p2 + 1, s1 + s2)
}
val widerType = widerDecimalType(p1, s1, p2, s2)
CheckOverflow(Multiply(promotePrecision(e1, widerType), promotePrecision(e2, widerType)),
resultType)

case Divide(e1 @ DecimalType.Expression(p1, s1), e2 @ DecimalType.Expression(p2, s2)) =>
var intDig = min(DecimalType.MAX_SCALE, p1 - s1 + s2)
var decDig = min(DecimalType.MAX_SCALE, max(6, s1 + p2 + 1))
val diff = (intDig + decDig) - DecimalType.MAX_SCALE
if (diff > 0) {
decDig -= diff / 2 + 1
intDig = DecimalType.MAX_SCALE - decDig
val resultType = if (SQLConf.get.decimalOperationsAllowTruncat) {
// Precision: p1 - s1 + s2 + max(6, s1 + p2 + 1)
// Scale: max(6, s1 + p2 + 1)
val intDig = p1 - s1 + s2
val scale = max(DecimalType.MINIMUM_ADJUSTED_SCALE, s1 + p2 + 1)
val prec = intDig + scale
DecimalType.adjustPrecisionScale(prec, scale)
} else {
var intDig = min(DecimalType.MAX_SCALE, p1 - s1 + s2)
var decDig = min(DecimalType.MAX_SCALE, max(6, s1 + p2 + 1))
val diff = (intDig + decDig) - DecimalType.MAX_SCALE
if (diff > 0) {
decDig -= diff / 2 + 1
intDig = DecimalType.MAX_SCALE - decDig
}
DecimalType.bounded(intDig + decDig, decDig)
}
val resultType = DecimalType.bounded(intDig + decDig, decDig)
val widerType = widerDecimalType(p1, s1, p2, s2)
CheckOverflow(Divide(promotePrecision(e1, widerType), promotePrecision(e2, widerType)),
resultType)

case Remainder(e1 @ DecimalType.Expression(p1, s1), e2 @ DecimalType.Expression(p2, s2)) =>
val resultType = DecimalType.bounded(min(p1 - s1, p2 - s2) + max(s1, s2), max(s1, s2))
val resultType = if (SQLConf.get.decimalOperationsAllowTruncat) {
DecimalType.adjustPrecisionScale(min(p1 - s1, p2 - s2) + max(s1, s2), max(s1, s2))
} else {
DecimalType.bounded(min(p1 - s1, p2 - s2) + max(s1, s2), max(s1, s2))
}
// resultType may have lower precision, so we cast them into wider type first.
val widerType = widerDecimalType(p1, s1, p2, s2)
CheckOverflow(Remainder(promotePrecision(e1, widerType), promotePrecision(e2, widerType)),
resultType)

case Pmod(e1 @ DecimalType.Expression(p1, s1), e2 @ DecimalType.Expression(p2, s2)) =>
val resultType = DecimalType.bounded(min(p1 - s1, p2 - s2) + max(s1, s2), max(s1, s2))
val resultType = if (SQLConf.get.decimalOperationsAllowTruncat) {
DecimalType.adjustPrecisionScale(min(p1 - s1, p2 - s2) + max(s1, s2), max(s1, s2))
} else {
DecimalType.bounded(min(p1 - s1, p2 - s2) + max(s1, s2), max(s1, s2))
}
// resultType may have lower precision, so we cast them into wider type first.
val widerType = widerDecimalType(p1, s1, p2, s2)
CheckOverflow(Pmod(promotePrecision(e1, widerType), promotePrecision(e2, widerType)),
Expand All @@ -137,9 +176,6 @@ object DecimalPrecision extends TypeCoercionRule {
e2 @ DecimalType.Expression(p2, s2)) if p1 != p2 || s1 != s2 =>
val resultType = widerDecimalType(p1, s1, p2, s2)
b.makeCopy(Array(Cast(e1, resultType), Cast(e2, resultType)))

// TODO: MaxOf, MinOf, etc might want other rules
// SUM and AVERAGE are handled by the implementations of those expressions
}

/**
Expand Down Expand Up @@ -243,17 +279,43 @@ object DecimalPrecision extends TypeCoercionRule {
// Promote integers inside a binary expression with fixed-precision decimals to decimals,
// and fixed-precision decimals in an expression with floats / doubles to doubles
case b @ BinaryOperator(left, right) if left.dataType != right.dataType =>
(left.dataType, right.dataType) match {
case (t: IntegralType, DecimalType.Fixed(p, s)) =>
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: I feel it's more readable to just put the new cases for literal before these 4 cases.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

unfortunately this is not really feasible since we match on different thigs: here we match on left.dataType and right.dataType, while for literals we match on left and right

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

we can do

(left, right) match {
  case (l: Literal, r) => ...
  
  case (DecimalType.Expression(p, s), r @ IntergralType()) => ...
}

b.makeCopy(Array(Cast(left, DecimalType.forType(t)), right))
case (DecimalType.Fixed(p, s), t: IntegralType) =>
b.makeCopy(Array(left, Cast(right, DecimalType.forType(t))))
case (t, DecimalType.Fixed(p, s)) if isFloat(t) =>
b.makeCopy(Array(left, Cast(right, DoubleType)))
case (DecimalType.Fixed(p, s), t) if isFloat(t) =>
b.makeCopy(Array(Cast(left, DoubleType), right))
case _ =>
b
}
nondecimalLiteralAndDecimal(b).lift((left, right)).getOrElse(
nondecimalNonliteralAndDecimal(b).applyOrElse((left.dataType, right.dataType),
(_: (DataType, DataType)) => b))
}

/**
* Type coercion for BinaryOperator in which one side is a non-decimal literal numeric, and the
* other side is a decimal.
*/
private def nondecimalLiteralAndDecimal(
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this rule newly introduced?

Copy link
Contributor Author

@mgaido91 mgaido91 Dec 21, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, it is. If we don't introduce this, we have a failure in Hive compatibility tests, because Hive use the exact precision and scale needed by the literals, while we, before this change, were using conservative values for each type. For instance, if we have a select 123.12345 * 3, before this change 3 would have been interpreted as Decimal(10, 0), which is the type for integers. After the change, 3 would become Decimal(1, 0), as Hive does. This prevents from needing more precision that what is actually needed.

b: BinaryOperator): PartialFunction[(Expression, Expression), Expression] = {
// Promote literal integers inside a binary expression with fixed-precision decimals to
// decimals. The precision and scale are the ones needed by the integer value.
case (l: Literal, r) if r.dataType.isInstanceOf[DecimalType]
&& l.dataType.isInstanceOf[IntegralType] =>
b.makeCopy(Array(Cast(l, DecimalType.forLiteral(l)), r))
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What if we don't do this? Requiring more precision seems OK as now we allow precision lose.

Copy link
Contributor Author

@mgaido91 mgaido91 Jan 16, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

if we don't do this we have many test failure in spark-hive, because hive does so. Moreover, requiring more precision is not OK, since it leads to a useless loss of precision. Think of this example: you multiply a column which is DECIMAL(38, 18) by 2. If you don't do this, 2 is considered a DECIMAL(10, 0). According to the rules, the result should be DECIMAL(38 + 10 + 1, 18), which is out of range: then according to the rules it becomes DECIMAL(38, 7), leading to potentially loosing 11 digits of the fractional part. With this change, instead, the result would be DECIMAL(38 + 1 + 1, 18), which becomes DECIMAL(38, 16), safely having a much lower precision loss.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

makes sense

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Add this example as the code comment?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hive is also doing this?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

yes, Hive is doing so. That is the reason why I introduced the change (without it, we would have had test failures in spark hive). I will add this in the comment.

case (l, r: Literal) if l.dataType.isInstanceOf[DecimalType]
&& r.dataType.isInstanceOf[IntegralType] =>
b.makeCopy(Array(l, Cast(r, DecimalType.forLiteral(r))))
}

/**
* Type coercion for BinaryOperator in which one side is a non-decimal non-literal numeric, and
* the other side is a decimal.
*/
private def nondecimalNonliteralAndDecimal(
b: BinaryOperator): PartialFunction[(DataType, DataType), Expression] = {
// Promote integers inside a binary expression with fixed-precision decimals to decimals,
// and fixed-precision decimals in an expression with floats / doubles to doubles
case (t: IntegralType, DecimalType.Fixed(p, s)) =>
b.makeCopy(Array(Cast(b.left, DecimalType.forType(t)), b.right))
case (DecimalType.Fixed(_, _), t: IntegralType) =>
b.makeCopy(Array(b.left, Cast(b.right, DecimalType.forType(t))))
case (t, DecimalType.Fixed(_, _)) if isFloat(t) =>
b.makeCopy(Array(b.left, Cast(b.right, DoubleType)))
case (DecimalType.Fixed(_, _), t) if isFloat(t) =>
b.makeCopy(Array(Cast(b.left, DoubleType), b.right))
}

}
Original file line number Diff line number Diff line change
Expand Up @@ -58,7 +58,7 @@ object Literal {
case s: Short => Literal(s, ShortType)
case s: String => Literal(UTF8String.fromString(s), StringType)
case b: Boolean => Literal(b, BooleanType)
case d: BigDecimal => Literal(Decimal(d), DecimalType(Math.max(d.precision, d.scale), d.scale))
case d: BigDecimal => Literal(Decimal(d), DecimalType.fromBigDecimal(d))
case d: JavaBigDecimal =>
Literal(Decimal(d), DecimalType(Math.max(d.precision, d.scale), d.scale()))
case d: Decimal => Literal(d, DecimalType(Math.max(d.precision, d.scale), d.scale))
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -1048,6 +1048,16 @@ object SQLConf {
.booleanConf
.createWithDefault(true)

val DECIMAL_OPERATIONS_ALLOW_TRUNCAT =
Copy link
Member

@dongjoon-hyun dongjoon-hyun Jan 12, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

typo? TRUNCAT -> TRUNCATE?

buildConf("spark.sql.decimalOperations.allowTruncat")
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

allowTruncat -> allowTruncate?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

thanks. This is what @cloud-fan suggested. Actually we are not truncating, but rounding. Personally, I'd prefer allowPrecisionLoss, but I am fine with any value. WDYT @cloud-fan @gatorsmile?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sorry that was my typo... allowPrecisionLoss SGTM

.internal()
.doc("When true, establishing the result type of an arithmetic operation happens " +
"according to Hive behavior and SQL ANSI 2011 specification, ie. rounding the decimal " +
"part of the result if an exact representation is not possible. Otherwise, NULL is" +
"returned in those cases, as previously (default).")
.booleanConf
.createWithDefault(false)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

we should make it true by default as it's a more reasonable behavior and follows Hive/SQL standard.


val SQL_STRING_REDACTION_PATTERN =
ConfigBuilder("spark.sql.redaction.string.regex")
.doc("Regex to decide which parts of strings produced by Spark contain sensitive " +
Expand Down Expand Up @@ -1423,6 +1433,8 @@ class SQLConf extends Serializable with Logging {

def replaceExceptWithFilter: Boolean = getConf(REPLACE_EXCEPT_WITH_FILTER)

def decimalOperationsAllowTruncat: Boolean = getConf(DECIMAL_OPERATIONS_ALLOW_TRUNCAT)

def continuousStreamingExecutorQueueSize: Int = getConf(CONTINUOUS_STREAMING_EXECUTOR_QUEUE_SIZE)

def continuousStreamingExecutorPollIntervalMs: Long =
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -23,7 +23,7 @@ import scala.reflect.runtime.universe.typeTag

import org.apache.spark.annotation.InterfaceStability
import org.apache.spark.sql.AnalysisException
import org.apache.spark.sql.catalyst.expressions.Expression
import org.apache.spark.sql.catalyst.expressions.{Expression, Literal}


/**
Expand Down Expand Up @@ -117,6 +117,7 @@ object DecimalType extends AbstractDataType {
val MAX_SCALE = 38
val SYSTEM_DEFAULT: DecimalType = DecimalType(MAX_PRECISION, 18)
val USER_DEFAULT: DecimalType = DecimalType(10, 0)
val MINIMUM_ADJUSTED_SCALE = 6
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Before naming a conf, I need to understand the rule you are following. https://docs.microsoft.com/en-us/sql/t-sql/data-types/precision-scale-and-length-transact-sql The SQL Server only applies MINIMUM_ADJUSTED_SCALE for multiplication and division. However, in your impl, you are using it for all the BinaryArithmetic operators?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, I followed Hive's implementation which works like this and applies this 6 digits minimum to all operations. This means that SQLServer allows to round more digits than us in those cases, ie. we ensure at least 6 digits for the scale, while SQLServer doesn't.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@gatorsmile what about spark.sql.decimalOperations.mode which defaults to native and accepts also hive (and in future also sql2011 for throwing exception instead of returning NULL)?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

how about spark.sql.decimalOperations.allowTruncat? Let's leave the mode stuff to the type coercion mode.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We should make it an internal conf and remove it after some releases.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ok, I'll go with that, thanks @cloud-fan.


// The decimal types compatible with other numeric types
private[sql] val ByteDecimal = DecimalType(3, 0)
Expand All @@ -136,10 +137,52 @@ object DecimalType extends AbstractDataType {
case DoubleType => DoubleDecimal
}

private[sql] def forLiteral(literal: Literal): DecimalType = literal.value match {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this different than forType if applied on Literal.dataType?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

yes, please see my comment above for an example. Thanks.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

fromLiteral?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

since we have forType I used forLiteral to be coherent on the naming

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

but we also have fromDecimal...

case v: Short => fromBigDecimal(BigDecimal(v))
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can't we just use ShortDecimal, IntDecimal...?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No, please see my comments above.

case v: Int => fromBigDecimal(BigDecimal(v))
case v: Long => fromBigDecimal(BigDecimal(v))
case _ => forType(literal.dataType)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

  private[sql] def forType(dataType: DataType): DecimalType = dataType match {
    case ByteType => ByteDecimal
    case ShortType => ShortDecimal
    case IntegerType => IntDecimal
    case LongType => LongDecimal
    case FloatType => FloatDecimal
    case DoubleType => DoubleDecimal
  }

This list is incomplete. Is that possible, the input literal is Literal(null, NullType)

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this problem was present before this PR. Should we fix it here? Is this fix needed? I guess that if it would have been a problem, it would already have been reported.

}

private[sql] def fromBigDecimal(d: BigDecimal): DecimalType = {
DecimalType(Math.max(d.precision, d.scale), d.scale)
}

private[sql] def bounded(precision: Int, scale: Int): DecimalType = {
DecimalType(min(precision, MAX_PRECISION), min(scale, MAX_SCALE))
}

/**
* Scale adjustment implementation is based on Hive's one, which is itself inspired to
* SQLServer's one. In particular, when a result precision is greater than
* {@link #MAX_PRECISION}, the corresponding scale is reduced to prevent the integral part of a
* result from being truncated.
*
* This method is used only when `spark.sql.decimalOperations.allowTruncat` is set to true.
*
* @param precision
* @param scale
* @return
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

remove the above 3 lines

*/
private[sql] def adjustPrecisionScale(precision: Int, scale: Int): DecimalType = {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The logics in this adjustment function is also different from the MS SQL Server docs.

In multiplication and division operations we need precision - scale places to store the integral part of the result. The scale might be reduced using the following rules:
The resulting scale is reduced to min(scale, 38 – (precision-scale)) if the integral part is less than 32, because it cannot be greater than 38 – (precision-scale). Result might be rounded in this case.
The scale will not be changed if it is less than 6 and if the integral part is greater than 32. In this case, overflow error might be raised if it cannot fit into decimal(38, scale)
The scale will be set to 6 if it is greater than 6 and if the integral part is greater than 32. In this case, both integral part and scale would be reduced and resulting type is decimal(38,6). Result might be rounded to 6 decimal places or overflow error will be thrown if integral part cannot fit into 32 digits.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

sorry, but I think this is exactly the same which is described there. The implementation might seem doing different things but actually the result will be the same. They both take the min between 6 and the desired scale if the precision is not enough to represent the whole scale.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So the rule in document is

val resultPrecision = 38
if (intDigits < 32) { // This means scale > 6, as iniDigits = precision - scale and precision > 38
  val maxScale = 38 - intDigits
  val resultScale = min(scale, maxScale)
} else {
  if (scale < 6) {
    // can't round as scale is already small
    val resultScale = scale
  } else {
    val resltScale = 6
  }
}

I think this is a little different from the current rule

val minScaleValue = Math.min(scale, 6)
val resultScale = max(38 - intDigits, minScaleValue)

Think aboout the case iniDigits < 32, SQL server is min(scale, 38 - intDigits), we are 38 - intDigits

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@cloud-fan yes, but you have to keep in mind that we are doing so only when precision is > 38. With some simple math (given intDigits = precision - scale), SQL server is min(scale, scale + 38 - precision). Since we perform this operation only when precision is greater than 38, the second member is always the minimum. Which means that in such a case, SQL server behaves like us, ie. it takes always 38 - intDigits. When precision is < than 38, instead we return the input precision and scale, as SQL server does. We are just using the precision instead of the intDigits for the if.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ah i see, makes sense

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, this part is consistent.

// Assumptions:
// precision >= scale
// scale >= 0
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Use assert to make sure assumptions?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I can add it even though it is not needed... there is no way we can violate those constraints. If you believe it is better to use assert, I will do that.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

use assert for assumptions, not comments.

if (precision <= MAX_PRECISION) {
// Adjustment only needed when we exceed max precision
DecimalType(precision, scale)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Shouldn't we also prevent scale > MAX_SCALE?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this is prevented outside this function.

} else {
// Precision/scale exceed maximum precision. Result must be adjusted to MAX_PRECISION.
val intDigits = precision - scale
// If original scale less than MINIMUM_ADJUSTED_SCALE, use original scale value; otherwise
// preserve at least MINIMUM_ADJUSTED_SCALE fractional digits
val minScaleValue = Math.min(scale, MINIMUM_ADJUSTED_SCALE)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sounds like MAXIMUM_ADJUSTED_SCALE instead of MINIMUM_ADJUSTED_SCALE.

Copy link
Contributor Author

@mgaido91 mgaido91 Dec 21, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It is the MINIMUM_ADJUSTED_SCALE. We can't have a scale lower that that, even though we would need not to loose precision. Please see the comments above.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We can't have a scale lower that that...

Don't you get a scale lower than MINIMUM_ADJUSTED_SCALE by Math.min(scale, MINIMUM_ADJUSTED_SCALE)?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, sorry, my answer was very poor, I will rephrase. scale contains the scale which we need to represent the values without any precision loss. What we are doing here is saying that the lower bound for the scale is either the scale that we need to correctly represent the value or the MINIMUM_ADJUSTED_SCALE. After this, in the line below we state that the scale we will use is the max between the number of digits of the precision we don't need on the left of the dot and this minScaleValue: ie. even though in some cases we might need a scale higher than MINIMUM_ADJUSTED_SCALE, but the number of digits needed on the left on the dot would force us to have a scale lower than MINIMUM_ADJUSTED_SCALE, we enforce that we will maintain at least MINIMUM_ADJUSTED_SCALE. We can't let the scale be lower that this threshold, even though it would be needed to enforce that we don't loose digits on the left of the dot. Please refer also to the blog post I linked in the comment above for further (hopefully better) explanation.

val adjustedScale = Math.max(MAX_PRECISION - intDigits, minScaleValue)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sounds like Math.min?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It is max because we take either the scale which would prevent a loss of "space" for intDigits, ie. the part on the left of the dot, or the minScaleValue, which is the scale we are ensuring to provide at least.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This line needs some comments.


DecimalType(MAX_PRECISION, adjustedScale)
}
}

override private[sql] def defaultConcreteType: DataType = SYSTEM_DEFAULT

override private[sql] def acceptsType(other: DataType): Boolean = {
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -408,8 +408,8 @@ class AnalysisSuite extends AnalysisTest with Matchers {
assertExpressionType(sum(Divide(1.0, 2.0)), DoubleType)
assertExpressionType(sum(Divide(1, 2.0f)), DoubleType)
assertExpressionType(sum(Divide(1.0f, 2)), DoubleType)
assertExpressionType(sum(Divide(1, Decimal(2))), DecimalType(31, 11))
assertExpressionType(sum(Divide(Decimal(1), 2)), DecimalType(31, 11))
assertExpressionType(sum(Divide(1, Decimal(2))), DecimalType(22, 11))
assertExpressionType(sum(Divide(Decimal(1), 2)), DecimalType(26, 6))
assertExpressionType(sum(Divide(Decimal(1), 2.0)), DoubleType)
assertExpressionType(sum(Divide(1.0, Decimal(2.0))), DoubleType)
}
Expand Down
39 changes: 39 additions & 0 deletions sql/core/src/test/resources/sql-tests/inputs/decimals.sql
Original file line number Diff line number Diff line change
@@ -0,0 +1,39 @@
--
-- Licensed to the Apache Software Foundation (ASF) under one or more
-- contributor license agreements. See the NOTICE file distributed with
-- this work for additional information regarding copyright ownership.
-- The ASF licenses this file to You under the Apache License, Version 2.0
-- (the "License"); you may not use this file except in compliance with
-- the License. You may obtain a copy of the License at
--
-- http://www.apache.org/licenses/LICENSE-2.0
--
-- Unless required by applicable law or agreed to in writing, software
-- distributed under the License is distributed on an "AS IS" BASIS,
-- WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-- See the License for the specific language governing permissions and
-- limitations under the License.
--

-- tests for decimals handling in operations
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why create a new test file instead of adding more cases in decimalArithmeticOperations.sql?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

because that file was meant for the typeCoercion modes (eg. if we introduce a sql2011 mode which throws exception instead of returning NULL), while this is more generic about arithmetic operations' behavior.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

that file is under .../typeCoercion/native/, which is meant for the default behavior(native mode of type coercion). If we introduce a sql2001 mode, we will put a same file under .../typeCoercion/sql2001/

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I see. I'll merge this file into the other then. Thanks.

create table decimals_test(id int, a decimal(38,18), b decimal(38,18)) using parquet;

insert into decimals_test values(1, 100.0, 999.0), (2, 12345.123, 12345.123),
(3, 0.1234567891011, 1234.1), (4, 123456789123456789.0, 1.123456789123456789);

-- test decimal operations
select id, a+b, a-b, a*b, a/b from decimals_test order by id;

-- test operations between decimals and constants
select id, a*10, b/10 from decimals_test order by id;

-- use rounding instead of returning NULL, according to new Hive's behavior and SQL standard
set spark.sql.decimalOperations.allowTruncat=true;

-- test decimal operations
select id, a+b, a-b, a*b, a/b from decimals_test order by id;

-- test operations between decimals and constants
select id, a*10, b/10 from decimals_test order by id;

drop table decimals_test;
Loading