Add documentation for handling of NaNs

prestodb · Jun 6, 2024 · b673668 · b673668
1 parent 8493792
commit b673668
Showing 1 changed file with 10 additions and 0 deletions.
diff --git a/presto-docs/src/main/sphinx/language/types.rst b/presto-docs/src/main/sphinx/language/types.rst
@@ -65,12 +65,22 @@ Floating-Point
 A real is a 32-bit inexact, variable-precision implementing the
 IEEE Standard 754 for Binary Floating-Point Arithmetic.
 
+Presto strays from the IEEE standard when handling NaNs.
+In Presto, NaN is considered larger than any other value for
+all comparison and sorting operations. Additionally, NaN=NaN will
+be true for all equality and distinctness purposes.
+
 ``DOUBLE``
 ^^^^^^^^^^
 
 A double is a 64-bit inexact, variable-precision implementing the
 IEEE Standard 754 for Binary Floating-Point Arithmetic.
 
+Presto strays from the IEEE standard when handling NaNs.
+In Presto, NaN is considered larger than any other value for
+all comparison and sorting operations. Additionally, NaN=NaN will
+be true for all equality and distinctness purposes.
+
 Fixed-Precision
 ---------------