DataFrame → Signals

Shipping

df_to_signals

Plot every column against a chosen X column — emits one signal per remaining numeric column

Signature

Inputs

dfDataFramerequired— The table to expand. Every numeric column (except the chosen X) becomes a bundle channel.

Outputs

signalsSignalBundle— One channel per numeric column, sharing the resolved X / time axis.

Parameters

Key	Type	Default	Notes
`x_column`	string	0	Column used as the shared X axis (name or 0-based index). EXCLUDED from the output signals. Empty or -1 = use the row index. A single-numeric-column frame is always plotted against the row index.
`sample_rate`	float	1000.0Hz	Sampling rate used to synthesize the time axis when X is the row index (ignored when an X column is chosen).
`unit`	string		Physical unit applied to channels whose source column carries no unit in its header.

Description

DataFrame → Signals (df_to_signals) expands a table back into a plottable SignalBundle — the node form of "see the columns plotted against the first column". It is the inverse of df_from_signals and shares its heavy lifting with the file_import "Signal Bundle" output mode.

One column is chosen as the shared X / time axis via x_column (a name or 0-based index). That column is excluded from the outputs; every other numeric column becomes a bundle channel sampled at the X values. If x_column is empty or -1, X is the row index — a uniform ramp at sample_rate Hz — and every numeric column becomes a channel.

The resolver is forgiving: a selector that does not resolve to a real column simply falls back to the row index rather than erroring. There is also an $n = 1$ fallback — if excluding the named X column would leave no numeric channel (i.e. the frame has only one numeric column), that lone column is emitted anyway, plotted against the row index. So a single-column table always yields a one-channel bundle.

Columns carrying a unit in their header keep it; columns with no unit adopt the Fallback Unit param. The node is stateless. A non-DataFrame input is a type mismatch.

Mathematics

channels = {col_{j} : j \neq = x, col_{j} numeric}

t_{i} = {col_{x} [i], i / f_{s}, x resolved x = row index

n_{numeric} = 1 \Rightarrow that column \mapsto single channel over the row index

Examples

Three columns, X = column 0

A frame with columns t = [0, 1, 2], a = [1, 2, 3], b = [10, 20, 30] and x_column = 0 yields a 2-channel bundle: a and b, each sampled at $t = 0, 1, 2$ . Column t is consumed as the axis, not emitted.

Single column falls back to the index axis

A frame with one numeric column v = [5, 6, 7] and x_column = 0 cannot exclude v and still have a channel, so the $n = 1$ fallback emits v as a single channel over the row index $0, 1, 2$ at sample_rate — no "no numeric columns" error.

Applications

Plotting an imported CSV/TSV table: pick the time column as X and watch every measurement column as a trace.
Round-tripping the columnar toolkit back to the signal domain after select / filter / sort / slice edits.
Turning a computed parameter frame (e.g. a regression sweep) into channels for a strip-chart or FFT.
Rendering an index-only numeric grid as uniformly sampled signals by leaving x_column empty and setting a sample rate.

Neat

An unresolvable X selector degrades gracefully to the row index instead of failing, so a typo in a column name still produces a usable plot.

The n = 1 fallback guarantees a single-numeric-column frame is never dropped: it is re-plotted against the row index even when the user named it as the X axis.

It reuses io::dataframe_to_bundle — the very same conversion file_import uses for its Signal Bundle output — so canvas and import paths render identical bundles.

Known issues

Non-numeric (text/datetime-as-string) columns are silently skipped as channels; only numeric columns become signals.

When X is the row index the time axis is a synthetic uniform ramp at sample_rate — it does not reflect any real timestamp column unless you select one as X.