A name the engine owns takes the prefix of the class or base type of its argument, and functions that belong together take the same prefix: spanLowerInc goes with spanLower, geoCumulativeLength with geoLength. Spark and Flink carry one set of names: each registers such a function under its prefixed name alone, so a query written for one runs on the other, and round(1.23456, 2) remains the engine's own rounding of a number. The names that change are the following.

PostgreSQL reads and writes a value through the four functions its type declares, and calls them implicitly: a typed literal, a cast from text and a text COPY go through the input function, the text a client prints through the output function, and a binary COPY or a driver in binary mode through the receive and send functions. Spark and Flink call none of them: a value travels as its WKB bytes, and each direction is a function the query names, the inverse of another, as the following table shows.

Every type has each pair it can state, the E forms when its text and binary forms omit the SRID it carries and MF-JSON when it is temporal, the types PostgreSQL reads and writes only through its four functions included: the set, span and span set types of integers, big integers, floats, dates and timestamps, textset and jsonbset, tbox and stbox, tbool, tint, tbigint, tfloat and ttext, and quadbin, s2cell, nsegment and tpcbox. A type whose text is its hex WKB, raquet, reads and writes its text through raquetFromHexWKB and asHexWKB, as a PostGIS raster through ST_RastFromHexWKB and ST_AsHexWKB. A box, stbox or tpcbox, writes its SRID in its text and its WKB, so it needs no E form. Every WKB output takes the byte order endian, and every text output of a value carrying floats or coordinates the precision maxdecimaldigits. A column of WKB bytes, a TemporalParquet column for instance, thus enters every engine through the binary constructor of its type, as in tgeompointFromBinary(col).

A function returning a set of rows, such as timeSplit or unnest, is called as a table function in the FROM clause in PostgreSQL. In Flink and in Spark the same function, unnest under the names setUnnest and temporalUnnest, returns an array, one element per row, an array of records when it returns several columns; Flink unfolds it into rows with CROSS JOIN UNNEST, and Spark with LATERAL VIEW explode, or inline for an array of records.

The query

SELECT vehicle, round(length(trip)::numeric, 1), lower(timeSpan(trip))
FROM Trips
WHERE trip && stbox 'SRID=3857;STBOX X((0,0),(1000,1000))';

is written, in Spark and Flink,

SELECT vehicle, round(geoLength(trip), 1), spanLower(timeSpan(trip))
FROM Trips
WHERE stboxOverlaps(trip, stboxFromText('SRID=3857;STBOX X((0,0),(1000,1000))'));

where round of the resulting number is the engine's own.