Struct PyRecords

pub struct PyRecords {
    inner: Records,
}

Expand description

A table of results: named columns of equal length, each carrying its kind.

Returned by every reading taken against market data — a valuation, a calibration, a fit, an attribution. A listing read off passive contract terms with no market in hand stays a list of tuples.

Fields§

§inner: Records

Implementations§

§

impl PyRecords

pub(crate) fn new(inner: impl Into<Records>) -> Self

This table from the core’s own record set.

fn column(&self, name: &str) -> Option<&Column>

The column named name, if the table holds one.

§

impl PyRecords

fn columns(&self) -> Vec<&'static str>

The column names, in the order the producing frame wrote them.

fn column_kinds(&self) -> Vec<&'static str>

Each column’s kind — "text", "number" or "date" — in column order.

What a caller building their own frame needs, and what the two dataframe readings declare their schema from.

fn len(&self) -> usize

How many rows the table holds.

Rows, while [...] takes a column name — the two axes the table has, and the pair that makes for row in records ambiguous. See __iter__, which refuses it rather than choosing.

fn getitem( &self, py: Python<'_>, key: &Bound<'_, PyAny>, ) -> PyResult<PyObject>

One column’s values, by name.

The key is taken as &Bound<PyAny> rather than &str so that an index of the wrong type is refused here. Left to the extractor it came back as “argument ‘name’: ‘int’ object cannot be converted to ‘PyString’”, which names neither the type being indexed nor the thing to do instead — and which is what a caller writing for row in records actually saw, because the sequence protocol reaches this with 0.

fn iter(&self) -> PyResult<PyObject>

Refused, and deliberately.

A table has two axes and the name says nothing about which one a loop wants. The two dataframe libraries a caller is likely to arrive from do not agree either — pandas iterates column names, polars iterates columns — so any choice here silently surprises half of them.

Without this the loop still failed, because __len__ made the sequence protocol apply and it reached __getitem__ with 0. It failed with a message about converting an int to a PyString, which sent at least one reader looking for a string they had passed wrongly. This says what to call instead.

fn repr(&self) -> String

fn str(&self) -> String

The table rendered for reading: a header, a rule, one line per row.

print(records) is the whole of the reading a caller at a prompt wants, so it needs no method. Text and dates sit left, numbers right with thousands separators; a numeric column prints two decimals when its largest magnitude is at least a thousand and six otherwise, so money reads as money and a rate or a discount factor keeps its digits. repr stays the one-line summary, which is what a traceback or a list of tables can bear.

fn _repr_html_(&self) -> String

The table as HTML, which is what a notebook shows for a cell whose value is a Records: the same cells str prints, numbers sitting right, text escaped, captioned with the row count.

fn to_dict<'py>(&self, py: Python<'py>) -> PyResult<Bound<'py, PyDict>>

The table as a mapping of column name to list of values.

Text becomes str, Number float, and Date datetime.date with None where a row has none. An absent number is nan, not None: the cell belongs to a numeric column and keeping it numeric is what stops the column changing type around it.

fn to_rows<'py>(&self, py: Python<'py>) -> Bound<'py, PyList>

The table as a list of row tuples, each in column order.

fn to_polars(&self, py: Python<'_>) -> PyResult<PyObject>

The table as a polars.DataFrame, built with a declared schema.

§Errors

ImportError naming the helixio[polars] extra when polars is absent.

fn to_pandas(&self, py: Python<'_>) -> PyResult<PyObject>

The table as a pandas.DataFrame, with declared dtypes.

A date column becomes datetime64[ns] with NaT where absent, rather than the object column pandas infers from datetime.date values.

§Errors

ImportError naming the helixio[pandas] extra when pandas is absent.

fn to_csv(&self, path: &str) -> PyResult<()>

Writes the table to path as CSV — a header row, then one line per row.

Imports nothing. Writing a text format through a dataframe library would make a plain to_csv the one call in this package that needs a dependency, which is what this avoids.