psi.data package

Subpackages

• psi.data.sinks package
  • Submodules
  • psi.data.sinks.api module
  • Module contents

Submodules

psi.data.api module

psi.data.plot_groups module

Grouped-epoch accumulation for the plotting subsystem.

EpochGroupAccumulator tracks epochs grouped by a (tab_key, plot_key) pair and implements the n_update batching rule: a group’s plot is only redrawn once at least n_update new epochs have arrived since its last redraw.

Averaging is incremental: each epoch is folded into a per-group running mean as it arrives (optionally through a transform applied per epoch, for plots that average in a transformed space such as dB-PSD). This keeps redraw cost proportional to the epoch length rather than to the number of epochs acquired so far, and means raw epochs do not need to be retained — memory is O(groups x samples) instead of O(total epochs).

Pure Python — no Qt — so the bookkeeping is testable without a GUI.

class psi.data.plot_groups.EpochGroupAccumulator(n_update=1, transform=None)

Bases: object

add_epochs(epochs, group_key)

Fold each epoch into the running mean for group_key(epoch.metadata).

Epochs whose key is None are excluded. Returns the key of the last epoch processed (which may be None), or None if epochs is empty.

count(key)

get_mean(key)

Running mean of the (transformed) epochs for this group, or None if the group has no epochs.

keys_for_tab(tab_key)

Return all group keys whose tab component matches tab_key.

mark_updated(key)

property max_samples

Longest epoch (in samples) seen across all groups (0 if no epochs).

needs_update(key)

reset()

tab_needs_update(tab_key)

True if at least one group in the tab has accumulated enough new epochs to warrant a redraw.

transform

Applied to each epoch before folding it into the group mean (e.g., dB-PSD for FFT-averaged plots). Must be a pure function of the epoch; parameters it closes over (fs, channel, …) must not change once epochs have been folded. Defaults to np.asarray.

class psi.data.plot_groups.RunningMean

Bases: object

Incremental mean of arrays, ignoring NaN values.

Arrays may grow along the last axis (ragged epochs): shorter arrays are treated as missing data (NaN) for the trailing samples, so the mean at each sample position reflects only the epochs that covered it.

Thread-safe: epochs are typically folded from an acquisition thread while the GUI thread reads the mean for rendering.

add(arr)

property mean

Current mean, or None if nothing has been added. Positions never covered by any epoch are NaN.

psi.data.plot_ranges module

Data-range state machines for the plotting subsystem.

These track the currently-visible x-range of a plot container based on the data flowing through the attached sources. Pure Atom classes — no pyqtgraph/Qt — so the windowing semantics can be tested without a GUI.

class psi.data.plot_ranges.BaseDataRange

Bases: Atom

add_source(source)

container

A value which allows objects of a given type or types.

Values will be tested using the PyObject_TypeCheck C API call. This call is equivalent to type(obj) in cls.mro(). It is less flexible but can be faster than Instance. Use Instance when allowing you need a tuple of types or (abstract) types relying on custom __isinstancecheck__ and Typed when the value type is explicit.

If optional is True, the value of a Typed may be set to None, otherwise None is not considered as a valid value.

current_range

A member which allows tuple values.

If item validation is used, then assignment will create a copy of the original tuple before validating the items, since validation may change the item values.

data_received(data)

delay

A value of type float.

By default, ints and longs will be promoted to floats. Pass strict=True to the constructor to enable strict float checking.

span

A value of type float.

By default, ints and longs will be promoted to floats. Pass strict=True to the constructor to enable strict float checking.

class psi.data.plot_ranges.ChannelDataRange

Bases: BaseDataRange

add_source(source)

current_time

A value of type float.

By default, ints and longs will be promoted to floats. Pass strict=True to the constructor to enable strict float checking.

data_received(data)

track_sources

A member which allows list values.

Assigning to a list creates a copy. The orginal list will remain unmodified. This is similar to the semantics of the assignment operator on the C++ STL container classes.

class psi.data.plot_ranges.EpochDataRange

Bases: BaseDataRange

add_source(source)

data_received(data)

max_duration

A value of type float.

By default, ints and longs will be promoted to floats. Pass strict=True to the constructor to enable strict float checking.

psi.data.plot_util module

Pure-Python helpers for the plotting subsystem.

Nothing in this module may import pyqtgraph or Qt: it holds the numeric and formatting logic (decimation, tick formatting, color cycles) so it can be tested without a GUI.

psi.data.plot_util.decimate_extremes(data, downsample)

psi.data.plot_util.decimate_mean(data, downsample)

psi.data.plot_util.format_log_ticks(values, scale, spacing)

psi.data.plot_util.format_time(seconds, fmt='{H:02d}:{M:02d}:{S:02.0f}')

psi.data.plot_util.get_color_cycle(name, n)

psi.data.plot_util.get_freq(fs, duration)

psi.data.plot_util.prepare_decimated_curve(data, t, downsample, mode)

Prepare (x, y, plot_kw) arrays for rendering a channel trace.

Parameters:

• data (1D array) – Signal for the visible range (NaN-filled where no data exists yet).

• t (1D array) – Time axis matching data.

• downsample (int) – Decimation factor (pixels per sample); applied when > 1 and mode is not ‘none’.

• mode ({'extremes', 'mean', 'none'}) – Decimation strategy.

Returns:

• (x, y, plot_kw) with empty arrays when the visible range contains no

• data (i.e., the plot should be cleared), or None when x and y are

• inconsistent (the caller should skip this update entirely — this can

• happen transiently while buffers resize).

psi.data.plots module

psi.data.plugin module

Module contents