upstream/postgresql
1
0
Fork 0
mirror of https://github.com/postgres/postgres.git synced 2026-06-09 00:32:10 -04:00
Code Issues Projects Releases Packages Wiki Activity Actions

Improve overflow/underflow handling in regr_intercept().

Browse source 
As with corr() and regr_r2(), improve regr_intercept()'s final
function to cope with overflow/underflow in the final calculation.
Here, instead of using sqrt(), we use frexp() and ldexp() to recover,
if an overflow or underflow is detected, so that the multiplication
and division steps operate on normalised mantissas, and cannot
overflow or underflow.

As with 6498287696, and the previous commit improving regr_r2(), this
is arguably a bug fix, but given the lack of prior complaints, refrain
from back-patching.

Reported-by: Tom Lane <tgl@sss.pgh.pa.us>
Author: Dean Rasheed <dean.a.rasheed@gmail.com>
Reviewed-by: Chengpeng Yan <chengpeng_yan@outlook.com>
Discussion: https://postgr.es/m/33E01656-BB3B-46E9-A41F-24A01A7C35F4@outlook.com
This commit is contained in:
Dean Rasheed 2026-06-03 09:20:21 +01:00
parent d58ec50e0f
commit eb8e76e130
3 changed files with 51 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

39
src/backend/utils/adt/float.c
View file

@ -4010,7 +4010,8 @@ float8_regr_intercept(PG_FUNCTION_ARGS)
Sx,
Sxx,
Sy,
Sxy;
Sxy,
dy;
transvalues = check_float8_array(transarray, "float8_regr_intercept", 8);
N = transvalues[0];
@ -4029,7 +4030,41 @@ float8_regr_intercept(PG_FUNCTION_ARGS)
if (Sxx == 0)
PG_RETURN_NULL();
PG_RETURN_FLOAT8((Sy - Sx * Sxy / Sxx) / N);
/*
* The intercept is given by (Sy - dy) / N, where dy = Sx * Sxy / Sxx.
* However, when computing dy, the intermediate product Sx * Sxy might
* underflow or overflow. If so, we can recover by decomposing Sx, Sxy,
* and Sxx into normalized mantissa and integer power-of-two components,
* computing the corresponding components of dy, and then recomposing dy.
* We avoid doing this if Sx, Sxy, or Sxx are infinite or NaN, since the
* exponent returned by frexp() is unspecified in those cases (and the
* final result would be the same in any case).
*/
dy = Sx * Sxy / Sxx;
if ((dy == 0 || isinf(dy)) &&
!(isinf(Sx) || isinf(Sxy) || isinf(Sxx) ||
isnan(Sx) || isnan(Sxy) || isnan(Sxx)))
{
float8 m_Sx,
m_Sxy,
m_Sxx,
m_dy;
int n_Sx,
n_Sxy,
n_Sxx,
n_dy;
m_Sx = frexp(Sx, &n_Sx);
m_Sxy = frexp(Sxy, &n_Sxy);
m_Sxx = frexp(Sxx, &n_Sxx);
m_dy = m_Sx * m_Sxy / m_Sxx;
n_dy = n_Sx + n_Sxy - n_Sxx;
dy = ldexp(m_dy, n_dy);
}
PG_RETURN_FLOAT8((Sy - dy) / N);
}

12
src/test/regress/expected/aggregates.out
View file

@ -563,6 +563,18 @@ SELECT corr(1e100 + g * 1e95, 1e100 + g * 1e95),
1 | 1
(1 row)
SELECT regr_intercept(y, x) FROM (VALUES (-1e150, 0), (2e150, 3e150)) v(x, y);
regr_intercept
----------------
1e+150
(1 row)
SELECT regr_intercept(y, x) FROM (VALUES (-1e-131, 0), (2e-131, 3e-131)) v(x, y);
regr_intercept
----------------
1e-131
(1 row)
-- these examples pose definitional questions for NaN inputs,
-- which we resolve by saying that an all-NaN input column is not all equal
SELECT corr(g, 'NaN'), regr_r2(g, 'NaN') FROM generate_series(1, 30) g;

2
src/test/regress/sql/aggregates.sql
View file

@ -159,6 +159,8 @@ SELECT corr(1e-100 + g * 1e-105, 1e-100 + g * 1e-105),
SELECT corr(1e100 + g * 1e95, 1e100 + g * 1e95),
regr_r2(1e100 + g * 1e95, 1e100 + g * 1e95)
FROM generate_series(1, 2) g;
SELECT regr_intercept(y, x) FROM (VALUES (-1e150, 0), (2e150, 3e150)) v(x, y);
SELECT regr_intercept(y, x) FROM (VALUES (-1e-131, 0), (2e-131, 3e-131)) v(x, y);
-- these examples pose definitional questions for NaN inputs,
-- which we resolve by saying that an all-NaN input column is not all equal