SPQL

SPQL is Spiral's library of functions and data types for working with Table data, from string and date functions such as spql.str.upper and spql.dt.year to image decoding and tensor math. A function name means the same thing everywhere you can call it: inside a Scan next to storage, on Arrow batches your application holds, or in a DataFusion query. So a value computed in one place matches the value computed in the others, and you can move work between them without changing results.

This reference lists every function with its signatures. Each function is a scalar function: it computes one value from the values in one row. Enrichment expressions use the same functions. To see exactly which functions and argument types your native library provides, call Functions::list.

Read a signatureCopy Anchor Link

One function name can accept several combinations of argument types. Each combination is an overload, with its own parameter types, output type, and nullability. When you call a function, Spiral picks the overload that matches your argument types before any rows are read. If none matches, the call fails right away. SPQL never casts arguments implicitly, so add an explicit cast when your types differ.

Some parameters must be literal values, such as a format name or a unit. A signature shows a literal parameter as a placeholder value, such as 'method' or 1, instead of a name. An optional parameter has a ? suffix.

The type after -> is the result. A ? on a type means it can be null. type(value) preserves that argument's type and nullability. An inferred result depends on the argument types; a declared result is supplied by the caller. Notes below each signature explain conditional nullability and any refinement by literal arguments.

NamespacesCopy Anchor Link

A function's name starts with a namespace that groups related functions, for example:

Data typesCopy Anchor Link

Signatures use the type names on the data types page. Semantic types, such as vector(n) or an image, give a meaning to a storage layout. Each semantic type has its own type ID, so a function that expects a vector rejects an arbitrary fixed-size list with the same layout.

Call a functionCopy Anchor Link

• In a Scan, build the call with Expression::call. See Reading.
• On your own Arrow batches, evaluate the calls with Functions. See Functions.
• In SQL, register the functions with with_spql and quote the dotted name, as in "spql.str.upper"(camera). See DataFusion.

The expressions page explains where each expression runs.

Functions by domainCopy Anchor Link

Each page documents one domain. For each function, it gives the signatures, the parameter types, and the result. The data types page defines the type names that the signatures use.