Documentation
¶
Overview ¶
Package decimalfmt renders Arrow decimal128 values as exact fixed-point strings, shared by the Thrift (arrowbased) and kernel result paths so a DECIMAL renders identically regardless of backend. Keeping one implementation means a precision fix (e.g. databricks-sql-go#274) lands in both at once.
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func Append ¶
func Append(dst []byte, n decimal128.Num, scale int32) []byte
Append renders n at the given scale into dst and returns the extended slice, exactly as ExactString would but without forcing a string allocation — for callers that render many decimals into a reused buffer (e.g. a batch-scoped arena). The output is byte-for-byte identical to ExactString.
The magnitude is rendered from the unscaled integer with no heap allocation when it fits in a uint64 (the common case: |unscaled| < 2^64 covers every DECIMAL(≤18) and every DECIMAL(≤38) value whose magnitude stays under 2^64). A true 128-bit magnitude (DECIMAL(20+) with a very large value) falls back to math/big, matching the old renderer's output exactly.
func ExactString ¶
func ExactString(n decimal128.Num, scale int32) string
ExactString renders an Arrow decimal128 as an exact fixed-point string, applying scale by digit placement rather than float conversion. This preserves precision a float64 would lose beyond ~17 significant digits (databricks-sql-go#274). A negative scale is not produced by the server and is treated as scale 0 rather than panicking.
It allocates only the returned string (the digit rendering uses a stack scratch, no math/big for magnitudes that fit in a uint64 — i.e. every DECIMAL(≤18) and most DECIMAL(≤38) values). See Append for the zero-copy form.
Types ¶
This section is empty.
Source Files