Structs Are Records

A struct declares named fields of any mix of types:

1
2
3
4
5
6

struct Trade [symbol, qty, price]

t = Trade['AAPL', 100, 191.5]
log[t.symbol]
log[t.qty * t.price]
log[t]

AAPL
19150
Trade[price = 191.5, qty = 100, symbol = 'AAPL']

Construct positionally (or with names — Trade[symbol = 'AAPL', ...]), access with dots. Structs pass to functions efficiently and print themselves readably.

Field names are free — schema names like type, sum, or count are fine as fields (t.type) even where they'd be reserved as variables. The only names a field can't take are the four error-metadata names every value already carries: ok, reason, kind, trace.

The decision rule, one line: same type under many keys → map; mixed types under fixed names → struct; a column of records → chapter 10's data frames.

Next: the loop that processes all of these at once.