PERCENTILE_CONT and PERCENTILE_DISC functions #8807

sim1984 · 2025-11-21T06:59:28Z

PERCENTILE_DISC and PERCENTILE_CONT functions

The PERCENTILE_CONT and PERCENTILE_DISC functions are known as inverse distribution functions.
These functions operate on an ordered set. Both functions can be used as aggregate or window functions.

PERCENTILE_DISC

PERCENTILE_DISC is an inverse distribution function that assumes a discrete distribution model.
It takes a percentile value and a sort specification and returns an element from the set.
Nulls are ignored in the calculation.

Syntax for the PERCENTILE_DISC function as an aggregate function.

PERCENTILE_DISC(<percent>) WITHIN GROUP (ORDER BY <expr> [ASC | DESC])

Syntax for the PERCENTILE_DISC function as an window function.

PERCENTILE_DISC(<percent>) WITHIN GROUP (ORDER BY <expr> [ASC | DESC])
  OVER (PARTITION BY <part_expr>)

The first argument <percent> must evaluate to a numeric value between 0 and 1, because it is a percentile value.
This expression must be constant within each aggregate group.
The ORDER BY clause takes a single expression that can be of any type that can be sorted.

The function PERCENTILE_DISC returns a value of the same type as the argument in ORDER BY.

For a given percentile value P, PERCENTILE_DISC sorts the values of the expression in the ORDER BY clause and
returns the value with the smallest CUME_DIST value (with respect to the same sort specification)
that is greater than or equal to P.

Analytic Example

SELECT 
  DEPT_NO, 
  SALARY,
  CUME_DIST() OVER(PARTITION BY DEPT_NO ORDER BY SALARY) AS "CUME_DIST",
  PERCENTILE_DISC(0.5) WITHIN GROUP(ORDER BY SALARY) 
    OVER(PARTITION BY DEPT_NO) AS MEDIAN_DISC
FROM EMPLOYEE
WHERE DEPT_NO < 600
ORDER BY 1, 2;

DEPT_NO                SALARY               CUME_DIST           MEDIAN_DISC
======= ===================== ======================= =====================
000                  53793.00      0.5000000000000000              53793.00
000                 212850.00       1.000000000000000              53793.00
100                  44000.00      0.5000000000000000              44000.00
100                 111262.50       1.000000000000000              44000.00
110                  61637.81      0.5000000000000000              61637.81
110                  68805.00       1.000000000000000              61637.81
115                6000000.00      0.5000000000000000            6000000.00
115                7480000.00       1.000000000000000            6000000.00
120                  22935.00      0.3333333333333333              33620.63
120                  33620.63      0.6666666666666666              33620.63
120                  39224.06       1.000000000000000              33620.63
121                 110000.00       1.000000000000000             110000.00
123                  38500.00       1.000000000000000              38500.00
125                  33000.00       1.000000000000000              33000.00
130                  86292.94      0.5000000000000000              86292.94
130                 102750.00       1.000000000000000              86292.94
140                 100914.00       1.000000000000000             100914.00
180                  42742.50      0.5000000000000000              42742.50
180                  64635.00       1.000000000000000              42742.50

PERCENTILE_CONT

PERCENTILE_CONT is an inverse distribution function that assumes a continuous distribution model.
It takes a percentile value and a sort specification and returns an element from the set.
Nulls are ignored in the calculation.

Syntax for the PERCENTILE_CONT function as an aggregate function.

PERCENTILE_CONT(<percent>) WITHIN GROUP (ORDER BY <expr> [ASC | DESC])

Syntax for the PERCENTILE_CONT function as an window function.

PERCENTILE_CONT(<percent>) WITHIN GROUP (ORDER BY <expr> [ASC | DESC])
  OVER (PARTITION BY <part_expr>)

The first argument <percent> must evaluate to a numeric value between 0 and 1, because it is a percentile value.
This expression must be constant within each aggregate group.
The ORDER BY clause takes a single expression, which must be of numeric type to perform interpolation.

The PERCENTILE_CONT function returns a value of type DOUBLE PRECISION or DECFLOAT(34) depending on the type
of the argument in the ORDER BY clause. A value of type DECFLOAT(34) is returned if ORDER BY contains
an expression of one of the types INT128, NUMERIC(38, x) or DECFLOAT(16 | 34), otherwise - DOUBLE PRECISION.

The result of PERCENTILE_CONT is computed by linear interpolation between values after ordering them.
Using the percentile value (P) and the number of rows (N) in the aggregation group, you can compute
the row number you are interested in after ordering the rows with respect to the sort specification.
This row number (RN) is computed according to the formula RN = (1 + (P * (N - 1)).
The final result of the aggregate function is computed by linear interpolation between the values from rows
at row numbers CRN = CEILING(RN) and FRN = FLOOR(RN).

function f(N) ::= value of expression from row at N

if (CRN = FRN = RN) then
  return f(RN)
else
  return (CRN - RN) * f(FRN) + (RN - FRN) * f(CRN)

Analytic Example

SELECT 
  DEPT_NO, 
  SALARY,
  PERCENT_RANK() OVER(PARTITION BY DEPT_NO ORDER BY SALARY) AS "PERCENT_RANK",
  PERCENTILE_CONT(0.5) WITHIN GROUP(ORDER BY SALARY) 
    OVER(PARTITION BY DEPT_NO) AS MEDIAN_CONT
FROM EMPLOYEE
WHERE DEPT_NO < 600
ORDER BY 1, 2;

DEPT_NO                SALARY            PERCENT_RANK             MEDIAN_CONT
======= ===================== ======================= =======================
000                  53793.00       0.000000000000000       133321.5000000000
000                 212850.00       1.000000000000000       133321.5000000000
100                  44000.00       0.000000000000000       77631.25000000000
100                 111262.50       1.000000000000000       77631.25000000000
110                  61637.81       0.000000000000000       65221.40500000000
110                  68805.00       1.000000000000000       65221.40500000000
115                6000000.00       0.000000000000000       6740000.000000000
115                7480000.00       1.000000000000000       6740000.000000000
120                  22935.00       0.000000000000000       33620.63000000000
120                  33620.63      0.5000000000000000       33620.63000000000
120                  39224.06      0.2500000000000000       33620.63000000000
121                 110000.00       0.000000000000000       110000.0000000000
123                  38500.00       0.000000000000000       38500.00000000000
125                  33000.00       0.000000000000000       33000.00000000000
130                  86292.94       0.000000000000000       94521.47000000000
130                 102750.00       1.000000000000000       94521.47000000000
140                 100914.00       0.000000000000000       100914.0000000000
180                  42742.50       0.000000000000000       53688.75000000000
180                  64635.00       1.000000000000000       53688.75000000000

An example of using both aggregate functions

SELECT 
  DEPT_NO, 
  PERCENTILE_CONT(0.5) WITHIN GROUP(ORDER BY SALARY) AS MEDIAN_CONT,
  PERCENTILE_DISC(0.5) WITHIN GROUP(ORDER BY SALARY) AS MEDIAN_DISC
FROM EMPLOYEE 
GROUP BY DEPT_NO;

DEPT_NO             MEDIAN_CONT           MEDIAN_DISC
======= ======================= =====================
000           133321.5000000000              53793.00
100           77631.25000000000              44000.00
110           65221.40500000000              61637.81
115           6740000.000000000            6000000.00
120           33620.63000000000              33620.63
121           110000.0000000000             110000.00
123           38500.00000000000              38500.00
125           33000.00000000000              33000.00
130           94521.47000000000              86292.94
140           100914.0000000000             100914.00
180           53688.75000000000              42742.50
600           66450.00000000000              27000.00
621           71619.75000000000              62550.00
622           53167.50000000000              53167.50
623           60000.00000000000              60000.00
670           71268.75000000000              31275.00
671           81810.19000000000              81810.19
672           45647.50000000000              35000.00
900           92791.31500000000              69482.63

…tions.

asfernandes · 2025-11-26T00:37:21Z

src/dsql/AggNodes.cpp

+		ERR_post(Arg::Gds(isc_wronumarg)); // WITHIN GROUP should only 1 ORDER item
+	}
+
+


Suggested change

asfernandes · 2025-11-26T00:37:55Z

src/dsql/AggNodes.cpp

+	fb_assert(dsqlOrderClause);
+	if (dsqlOrderClause->items.getCount() != 1)
+	{
+		ERR_post(Arg::Gds(isc_wronumarg)); // WITHIN GROUP should only 1 ORDER item


This error has margin to deduce different things.

asfernandes · 2025-11-26T00:42:56Z

src/dsql/AggNodes.cpp

+	AggNode::pass2(tdbb, csb);
+
+	// We need a second descriptor in the impure area for PERCENTILE.
+	impure2Offset = csb->allocImpure<impure_value_ex>();


Why did you need to extra impure area instead of declare a single struct with all (and only) the things you need?

asfernandes · 2025-11-26T00:45:40Z

src/dsql/AggNodes.cpp

+	if (desc->isNull())
+		return false;
+


This should not be necessary.

asfernandes · 2025-11-26T00:46:24Z

src/dsql/AggNodes.cpp

+		if (desc->isBlob())
+			ERRD_post(Arg::Gds(isc_blobnotsup) << Arg::Str("ORDER BY"));		
+
+		const auto percentile_value = MOV_get_double(tdbb, percenteDesc);


Name things in the more modern convention, please: percentileValue

asfernandes · 2025-11-26T00:47:03Z

src/dsql/AggNodes.cpp

+		for (auto& nodeOrder : sort->expressions)
+		{
+			dsc toDesc = *(descOrder++);
+			toDesc.dsc_address = data + (IPTR)toDesc.dsc_address;


Suggested change

toDesc.dsc_address = data + (IPTR)toDesc.dsc_address;

toDesc.dsc_address = data + (IPTR) toDesc.dsc_address;

asfernandes · 2025-11-26T00:48:08Z

src/dsql/AggNodes.cpp

+			EVL_make_value(tdbb, desc, impure2);
+		}
+	}
+	else {


Suggested change

else {

else

{

asfernandes · 2025-11-26T00:49:31Z

src/dsql/AggNodes.cpp

+					impure2->vlu_misc.vlu_double += value * (percentileImpure->crn - percentileImpure->rn);
+				}
+			}
+			if (impure2->vlux_count == percentileImpure->crn)


Suggested change

if (impure2->vlux_count == percentileImpure->crn)

if (impure2->vlux_count == percentileImpure->crn)

asfernandes · 2025-11-26T01:04:57Z

src/dsql/AggNodes.h

+
+private:
+	const PercentileType type;
+	NestConst<ValueExprNode> valueArg;


Don't you need to override getChildren and add these values?

sim1984 added 5 commits November 20, 2025 19:54

Percentile functions

398a9f5

Optimization. Record numbers are calculated only once.

732db5a

Fixed PERCENTILE_DISC with non-numeric argument

314378c

Getting the parameter type when preparing a query for percentile func…

a1f25a5

…tions.

update doc

5179bfe

asfernandes reviewed Nov 26, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

PERCENTILE_CONT and PERCENTILE_DISC functions #8807

PERCENTILE_CONT and PERCENTILE_DISC functions #8807

Uh oh!

sim1984 commented Nov 21, 2025 •

edited

Loading

Uh oh!

asfernandes Nov 26, 2025

Uh oh!

asfernandes Nov 26, 2025

Uh oh!

asfernandes Nov 26, 2025

Uh oh!

asfernandes Nov 26, 2025

Uh oh!

asfernandes Nov 26, 2025

Uh oh!

asfernandes Nov 26, 2025

Uh oh!

asfernandes Nov 26, 2025

Uh oh!

asfernandes Nov 26, 2025

Uh oh!

asfernandes Nov 26, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		ERR_post(Arg::Gds(isc_wronumarg)); // WITHIN GROUP should only 1 ORDER item
		}

	toDesc.dsc_address = data + (IPTR)toDesc.dsc_address;
	toDesc.dsc_address = data + (IPTR) toDesc.dsc_address;

	if (impure2->vlux_count == percentileImpure->crn)

	if (impure2->vlux_count == percentileImpure->crn)

Uh oh!

PERCENTILE_CONT and PERCENTILE_DISC functions #8807

Are you sure you want to change the base?

PERCENTILE_CONT and PERCENTILE_DISC functions #8807

Uh oh!

Conversation

sim1984 commented Nov 21, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

PERCENTILE_DISC and PERCENTILE_CONT functions

PERCENTILE_DISC

Analytic Example

PERCENTILE_CONT

Analytic Example

An example of using both aggregate functions

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

sim1984 commented Nov 21, 2025 •

edited

Loading