Nadav Har'El created CASSANDRA-20723:
----------------------------------------
Summary: decimal literal parsed as as a double, with misleading
error message
Key: CASSANDRA-20723
URL: https://issues.apache.org/jira/browse/CASSANDRA-20723
Project: Apache Cassandra
Issue Type: Bug
Reporter: Nadav Har'El
According to the Cassandra documentation, the CQL "decimal" type is based on
Java's BigDecimal type. The documentation of that type explains that the
so-called "scale" of this number (fractional digits minus exponent) can be very
large, and in particular, for example, the number "1e2147483647" (the exponent
is 2^31-1) should be a valid decimal.
However, experimentally if I create a table with a "decimal" clustering key and
try to insert any exponent 309 or higher, I get an error:
{{CREATE TABLE tbl (p int, c decimal, PRIMARY KEY (p, c))}}
{{INSERT INTO tbl (p, c) VALUES (1, 1e309)}}
The error I get is:
{{cassandra.protocol.SyntaxException: <Error from server: code=2000 [Syntax
error in CQL query] message="Failed parsing statement: [INSERT INTO tbl (p, c)
VALUES (1, 1e309)] reason: NumberFormatException Character I is neither a
decimal digit number, decimal point, nor "e" notation exponential mark.">}}
This error message is silly and misleading - which "Character I" does it refer
to? But beyond the wrong error message, it is also wrong to be an error at all,
because as I explained above 1e309 should have been a valid decimal value -
even 1e2147483647 is.
Since 1e308 is accepted but 1e309 is not, I suspect the bug is that the decimal
literal is parsed as a double-precision number and then assigned into a
decimal. But this is wrong - it means that with inline literals (without
prepared statements) we cannot actually use the full range of the decimal type:
We cannot use the full range of the exponent, and probably also can't use
arbitrary precision (but I didn't test this).
Is there a different way to specify a decimal literal? The documentation
doesn't seem to say anything about this question. I tried maybe a string like
"1e309" might work to initialize a decimal, but no - it doesn't seem to work.
This seems to be a *regression in Cassandra 4:* I tested Cassandra 3.11.17, and
it doesn't have this bug: It is able able to parse 1e309 and even 1e2147483647
just fine and assign them into the decimal clustering-key in my test. But the
bug does exist on Cassandra 4.1.6 and 5.0.0 which I tested.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
