Results and types

Results

execute returns a Result:

Member	Contents
`result.columns`	one `ColumnInfo` per column: name, declared type, nullability, precision, scale, and the length of a text or binary column
`result.values`	every cell, aligned with `columns`
`result.updatecount`	rows affected by DML, or `-1` for a query
`result.counters`	the raw `number of rows …` counters behind `updatecount`
`rows(result)`	each row as a dictionary keyed by column name, built on first use
`scalar(result)`	the first cell of the first row
`rowcount(result)`	rows returned, or rows affected by DML
`isupdate(result)`	whether the result came from DML
`columnnames(result)`	the column names, in order

A Result iterates and indexes over its rows:

for row in result
    println(row[1], " ", row[2])
end

result[1]           # the first row
result[1, 2]        # one cell of it
result[1, "NAME"]   # the same, by column name (case-insensitive)
map(row -> row[1], result)

rows cannot hold two columns with the same name (a self-join reports ID twice, and the later one wins); values keeps both.

Types

SQL type	Julia type
integral `NUMBER`, `INTEGER` and similar	`Int64`, or `BigInt` beyond 64 bits
fractional `NUMBER`, `FLOAT`, `DOUBLE`, `REAL`	`Float64`
`VARCHAR` and other text types	`String`
`BOOLEAN`	`Bool`
`BINARY`	`Vector{UInt8}`
`DATE`	`Date`
`TIME`	`Time`
`TIMESTAMP`, `TIMESTAMP_NTZ`, `DATETIME`	`DateTime`
`TIMESTAMP_LTZ`, `TIMESTAMP_TZ`	`ZonedTimestamp`
`VARIANT`, `OBJECT`, `ARRAY`	`String`, as the engine renders it
SQL `NULL`	`nothing`

A NUMBER(38,0) can exceed Int64. The driver's JSON parser keeps each number's text, so such a value arrives as an exact BigInt rather than a rounded Float64.

Julia's standard library has no zoned timestamp type, so TIMESTAMP_TZ and TIMESTAMP_LTZ become a ZonedTimestamp: the wall clock (z.datetime) and its UTC offset (z.offset). Compare two of them with utc(z), which returns the instant.

Date and DateTime values are read exactly as stored, without conversion to the host's time zone.