Struct RecordSource 

pub struct RecordSource<I> { /* private fields */ }

Rows from a stream of JSON records.

The source reads the first sample records ahead: their keys, in the order first seen, are the columns, and their values decide the schema (integer and float widen to float; disagreeing types become any; a key some records lack, or that holds null, is nullable). A record that is not an object is one value column. Scalars become cells as they are (true/false as text, since Value has no boolean; whole numbers in a float column as floats), and arrays and objects become their compact JSON text.

Keys first seen after the sample have no column; their values are dropped and their names kept in unknown_keys, so a caller can say so: the first MAX_UNKNOWN_KEYS of them, so a stream whose every record has a new key stays bounded (more_unknown_keys says there were more). Read everything at once (the adapters’ read) to sample every record.

Implementations

impl<I> RecordSource<I>

where I: Iterator<Item = Result<Value, DataError>>,

pub fn unknown_keys(&self) -> impl Iterator<Item = &str>

Keys seen after the sample, which have no column (so far): the first MAX_UNKNOWN_KEYS seen, sorted.

pub fn more_unknown_keys(&self) -> bool

Whether more keys than unknown_keys holds were seen after the sample.

Trait Implementations

impl<I: Debug> Debug for RecordSource<I>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<I> RowSource for RecordSource<I>

where I: Iterator<Item = Result<Value, DataError>>,

fn columns(&self) -> &[String]

The column names, in order.

fn schema(&self) -> Option<&Schema>

What the source knows about the columns’ types, if anything. A source’s schema has one field per column, in the same order.

fn next_row(&mut self) -> Option<Result<Row, DataError>>

The next row, None at the end, or why it could not be read.

fn collect_rows(self) -> Result<Rows, DataError>

where Self: Sized,

Read every remaining row into memory.