BUG #19101: Ceil on BIGINT could lost precision in decil function

On Sun, 2025-11-02 at 15:16 +0000, PG Bug reporting form wrote:

PostgreSQL version: 18.0

I try to store a large number in `BIGINT` and run `ceil(c1)` command.
However, the result lost some precision due to calling `decil` function.
```sql
CREATE TABLE t1 (c1 BIGINT);
INSERT INTO t1 VALUES (4854233034440979799);
-- dceil
SELECT ceil(c1) FROM t1; -- {4.854233034440979e+18}
```
The original number is expected to return.

This is not a bug. There are two ceil() functions:

List of functions
Schema │ Name │ Result data type │ Argument data types │ Type
════════════╪══════╪══════════════════╪═════════════════════╪══════
pg_catalog │ ceil │ double precision │ double precision │ func
pg_catalog │ ceil │ numeric │ numeric │ func

There are implicit casts from "bigint" to both "numeric" and "double precision":

List of casts
Source type │ Target type │ Function │ Implicit?
══════════════════╪══════════════════╪══════════╪═══════════════
...
bigint │ double precision │ float8 │ yes
...
bigint │ numeric │ numeric │ yes

There are two preferred numeric data types, and "numeric" is none of them:

SELECT typname FROM pg_type WHERE typcategory = 'N' AND typispreferred;

typname
═════════
oid
float8 (which is the same as "double precision")

Consequently, rule 4 d of the type conversion rules for function calls
(https://www.postgresql.org/docs/current/typeconv-func.html)
decrees that the "bigint" be case to "double precision", which explains
the rounding errors.

Use an explicit type cast:

SELECT ceil(c1::numeric) FROM t1;

Yours,
Laurenz Albe

PG Bug reporting form <noreply@postgresql.org> writes:

I try to store a large number in `BIGINT` and run `ceil(c1)` command.
However, the result lost some precision due to calling `decil` function.
```sql
CREATE TABLE t1 (c1 BIGINT);
INSERT INTO t1 VALUES (4854233034440979799);
-- dceil
SELECT ceil(c1) FROM t1; -- {4.854233034440979e+18}
```

This is not a bug. There are two versions of ceil() and you're
invoking the wrong one. You'd need to explicitly cast the
argument to numeric if you want ceil(numeric) to be used.

In this context it's a bit unfortunate that the parser's type
preference rules [1]https://www.postgresql.org/docs/current/typeconv.html prefer float8 to numeric. But we're pretty
much stuck with that behavior because (a) the SQL standard
says so [2]Well, what it really says is that expressions that mix exact and inexact numeric types produce inexact results. We interpret that as meaning that float8 is the preferred type in the numeric category, so it wins ambiguous cases., and (b) even if it didn't, we have a couple of
decades of history to be backwards compatible with.

regards, tom lane

[1]: https://www.postgresql.org/docs/current/typeconv.html

[2]: Well, what it really says is that expressions that mix exact and inexact numeric types produce inexact results. We interpret that as meaning that float8 is the preferred type in the numeric category, so it wins ambiguous cases.
exact and inexact numeric types produce inexact results.
We interpret that as meaning that float8 is the preferred
type in the numeric category, so it wins ambiguous cases.

Thanks for the quick response. I try to add an explicit type conversion to avoid this problem.
At 2025-11-03 00:35:26, "Tom Lane" <tgl@sss.pgh.pa.us> wrote: