Parsing error for small Float64 literals in scientific notation

[rienath]

Company or project name

ClickHouse

Describe what's wrong

Very small floating-point literals fail to parse when written in scientific notation. For example, 4.94066e-3 and 4.94066e-32 work fine, but 4.94066e-324 causes a syntax error. This seems to be a parsing issue as the value is a valid Float64.

Does it reproduce on the most recent release?

Yes

How to reproduce

SELECT 4.94066e-3; -- works
SELECT 4.94066e-32; -- works
SELECT 4.94066e-324; -- throws

Link to reproduce.

Expected behavior

The last query should return a very small Float64 value without triggering an error.

Error message and/or stacktrace

Code: 62. DB::Exception: Syntax error: failed at position 20 (end of query): ;. Expected one of: token, DoubleColon. (SYNTAX_ERROR)

[rienath]

@evillique maybe you could have a look? :)

[rschu1ze]

What is interesting is that 4.94066e-305 is still parseable, so the presence of three exponent digits isn't the problem.

https://fiddle.clickhouse.com/51d15975-8f3e-4ae4-8593-f8eefe593f12

The syntax docs (-1e-100) also have no issue with three digits.

[rienath]

Also interesting: the lowest parseable number is ≈ 2.2250738585072013e-308, which corresponds to the exponent part of the float being 00000000001 and the rest zeroed out.

1. https://fiddle.clickhouse.com/93788348-621e-41b7-ace3-a3fa3df7d024
2. https://www.binaryconvert.com/result_double.html?decimal=050046050050053048055051056053056053048055050048049051101045051048056

[Algunenano]

The fact that this number is literally DBL_MIN should hint it's likely some internal limit of whatever function we are using to parse the floating point value