The goal of this PR is to integrate the Decimal type (implemented in #5991) into the database and make it available to users through SQL, ILP and Parquet.

This integration will support decimal that have a precision of 76 or less, while using the relevant type to store the required value.

SQL

This PR will add the decimal column type, usable through CREATE TABLE, ALTER TABLE [table name] ADD COLUMN or ALTER TABLE [table name] ALTER COLUMN [column name] TYPE.

To declare a decimal column type, you need to use the syntax: DECIMAL(precision, scale).

• The precision defines the maximum count of digits that can be stored into this column. defaults to min(scale + 15, 76)
• The scale defines the maximum count of digits in the right side of the dot (a.k.a. the fractional part). defaults to 3

For example, to create a table with decimal columns:

CREATE TABLE trades (
  timestamp TIMESTAMP,
  symbol SYMBOL,
  amount DECIMAL(36, 8),
  price DECIMAL(18, 4)
) TIMESTAMP(timestamp)
PARTITION BY day;

This PR also allows you to do some basic arithmetic operations on decimals such as +, -, /, *, % and to compare decimal types.

ILP

This PR allows decimals to be inserted through the InfluxDB Line Protocol.

PGWire

With this PR, the decimal type should be compliant with the PGWire protocol, EXCEPT that floating point literals are interpreted as double and not decimal (unlike postgres).

As a consequence, when inserting a value to a decimal column (without the binary format), the user have to manually cast it to a decimal (we don't implicitly convert doubles to decimal to avoid unexpected precision loss) or use the 'm' suffix:

CREATE TABLE x (value DECIMAL);
INSERT INTO x VALUES (123.456); -- fails, double cannot be implicitly cast to decimal
INSERT INTO x VALUES (123.456m); -- ok: 'm' suffix -> we interpret the literal as a decimal
INSERT INTO x VALUES (cast(123.456 as DECIMAL)); -- ok: cast used, but might lead to precision loss

TODO

• Add null sentinel value for Decimal256.
• Implement memory I/O methods to Memory classes.
• Implement serialization in TableWriter.
• Implement in RowCopier.
• Implement in RecordSink.
• Implement in RecordValueSink. - @RaphDal
• Implement in BindVariableService.
• Implement in WalEventCursor.
• Add column metadata to store the precision/scale values.
• Check precision/scale values when storing decimals.
• Implement getDecimal[128/256] in Record.
• Add DECIMAL(p, s) syntax to SQL parser and lexer.
• Implement literal parsing in SQL.
• Add specific literal (e.g. 23m).
• Lossless casting from literal constant double.
• Implicit widening cast.
• Support CREATE TABLE.
• Support ALTER COLUMN TYPE. - @RaphDal
• Implement GROUP BY with Decimal values. - @RaphDal
• Implement arithmetic functions (add, subtract, multiply, divide, modulo, negate). - @puzpuzpuz
• Implement comparison operators (<=, <, ==, >, >=). - @puzpuzpuz
• Implement Decimal support in record comparators, so that ORDER BY works with decimal columns and expressions. - @RaphDal
• Implement type casting functions (to/from double, long, int, varchar, etc.). - @RaphDal
• Implicit cast from long/int/short/byte:
  • Functions
  • INSERT
  • INSERT AS SELECT - @RaphDal
  • UPDATE
• Support UPDATEs on Decimal columns.
• Support Decimal columns as dedup keys. - @RaphDal
• Support CSV import/export. - @RaphDal
• Support ILP. - @RaphDal
• Support PGWire. - @RaphDal
  • Binding binary format
• Support HTTP Rest API.
• Implement Decimal support in concat, coalesce, nullif. - @RaphDal
• Implement Decimal support in CASE.
• Implement rnd_decimal. - @puzpuzpuz
• Fix ::decimal(18,5) error (cast(x as decimal(18,5)) works).
• Implement basic aggregate functions for Decimal: min/max, avg, first/last, first_not_null/last_not_null, sum, count. - @puzpuzpuz
• Implement SAMPLE BY fill support for Decimal. - @RaphDal
  • [nice-to-have] LINEAR interpolation
• [nice-to-have] Implement window functions for Decimal: same set as for double.
• Implement basic math functions: abs, round*, sign, least, greatest, floor, ceil. - @RaphDal
• Update documentation to add the Decimal type. - @RaphDal
• [nice-to-have] Decimal support in Parquet.
• [nice-to-have] Optimize multiplication and division by pow of 10.
• [nice-to-have] Decimal operators support in JIT.
• [nice-to-have] Implement advanced aggregate functions for Decimal: count_distinct, corr, covar_pop, covar_samp, stddev/stddev_samp, stddev_pop, string_agg, string_distinct_agg, variance/var_samp, var_pop.
• [nice-to-have] Implement advanced math functions: ln, log, exp, sqrt.
• [nice-to-have] Rosti support for simple Decimal aggregate functions.

Outstanding test TODOs:

• Test INSERTs with Decimal values.
• Test GROUP BY with Decimal values.
• Test UNION with Decimal values, including mixed precisions/scales.
• Test JOINs with Decimal values, including implicit casts.
• Test Decimal coltops.
• Test Decimal bind variables.