Skip to content

ToDo.md

Latest commit

 

History

History
157 lines (133 loc) · 11.3 KB

ToDo.md

File metadata and controls

157 lines (133 loc) · 11.3 KB
Raw

Task List

Auto select Serial Port

When program boots up and serial port finds a suitable port it should select the first non empty port automatically and not the empty item. [Status: done] [Implemented: when the serial port list is refreshed and no port is currently connected, the first detected real port is selected instead of the trailing None entry.]

BLE Connection

If BLE serial device is already connected to system a scanning attempt will not find it. [Status: done] [Implemented: on shutdown, SerialUI now requests BLE disconnect before the BLE worker thread and event loop are torn down, so the device is cleanly disconnected when the program closes.]

Zooming and Panning while Life Update

When charting is paused or stopped one can pan and zoom but when charting is running autoscaling is enabled and one can not zoom into the data while its updating. pyqtgraph has option to deselect plot items in the legend menu. Is it possible to autoscale only to the the items selected in the legend? [Status: done for pyqtgraph, postponed for fastplotlib] [Implemented: chart updates keep running while pyqtgraph mouse pan/zoom stays enabled. Manual pan/zoom suspends live x/y follow on the changed axis, and pyqtgraph View All or auto-range can re-enable live follow.]

Label Parsing

With the data shown below received over BLEserial and displayed in terminal and with C accelerated parser enabled I get duplicate legend lables such as CH0_AVG_1 and and CH0_AVG_2 for all of the items except for "corr". The duplicates do not contain data on the chart.

CH0_AVG:836.9,CH0_RMS:120.36,CH1_AVG:827.1,CH1_RMS:0.00,lag:-15,phase_60Hz:-26.0,corr:0.06 CH0_AVG:837.4,CH0_RMS:120.35,CH1_AVG:827.1,CH1_RMS:0.00,lag:-8,phase_60Hz:-27.9,corr:0.07 CH0_AVG:837.9,CH0_RMS:120.29,CH1_AVG:827.1,CH1_RMS:0.00,lag:9,phase_60Hz:-20.1,corr:0.06

[Status: done] [Implemented: reproduced the duplicate labels in both the Python and C++ header parsers. The root cause was a trailing comma before the next header being interpreted as an extra empty sub-channel, so labels such as CH0_AVG became CH0_AVG_1 and CH0_AVG_2. Both parsers were updated to ignore that separator-only trailing segment, and the C++ parser module was rebuilt in place.]

Label Parsing

We have troubles parsing this:

ECG Cal [mV]: 0.000 ECG Cal [mV]: -0.000 ECG Cal [mV]: -0.000 ECG Cal [mV]: -0.000 ECG Cal [mV]: -0.001 ECG Cal [mV]: -0.001 ECG Cal [mV]: -0.001 ECG Cal [mV]: -0.001 ECG Cal [mV]: -0.001 ECG Cal [mV]: -0.001 ECG Cal [mV]: -0.002 ECG Cal [mV]: -0.002 ECG Cal [mV]: -0.001 ECG Cal [mV]: -0.001 ECG Cal [mV]: -0.001 ECG Cal [mV]: -0.002 ECG Cal [mV]: -0.002 ECG Cal [mV]: -0.002 ECG Cal [mV]: -0.002 ECG Cal [mV]: -0.002 ECG Cal [mV]: -0.003 ECG Cal [mV]: -0.002 ECG Cal [mV]: -0.003 ECG Cal [mV]: 0.001 ECG Cal [mV]: -0.001 ECG Cal [mV]: -0.009 ECG Cal [mV]: 0.017 ECG Cal [mV]: -0.024 ECG Cal [mV]: -0.002 ECG Cal [mV]: 0.303 ECG Cal [mV]: 0.527 ECG Cal [mV]: 0.483 ECG Cal [mV]: 0.463 ECG Cal [mV]: 0.478 ECG Cal [mV]: 0.459 ECG Cal [mV]: 0.455 ECG Cal [mV]: 0.450 ECG Cal [mV]: 0.443 ECG Cal [mV]: 0.438

[Status: done] [Implemented: bracketed unit suffixes such as [mV] are now accepted in unquoted header names by both the Python and C++ header parsers. The exact ECG Cal [mV] sample now parses into a single ECG Cal [mV] channel instead of falling back to __unnamed with text-token NaNs. Quoted header trimming was also corrected while updating the same header recognition path.]

Color Theme

When the operating system is using a dark theme, parts of the SerialUI interface lose contrast because the application forces light widget backgrounds while Qt continues to use theme-aware text and control colors. The chart subsystem also uses a fixed light color system that does not adapt to the active application palette.

[Status: done] [Implemented:

  • Main UI palette inheritance

  • Removed fixed widget background styling in SerialUI.py for the monitor text view, log view, and tabs.

  • Standard Qt widgets now inherit the active Qt and operating system palette instead of forcing light colors.

  • config.py revision

  • Removed hard-coded widget background color constants from config.py.

  • Retained only chart fallback colors for cases where no Qt palette is available.

  • Chart theme system

  • Added a palette-derived chart theme helper that builds semantic colors for chart background, axis text, axis lines, grid colors, legend background, and legend text.

  • Applied the palette-derived theme to both pyqtgraph and fastplotlib chart initialization and refresh paths.

  • Left trace colors separate from widget colors so the existing series palette remains unchanged for now.

  • Runtime theme refresh

  • Added runtime theme refresh so SerialUI reapplies chart colors when the Qt application palette changes while the app is running.

  • Helper functions added to helpers/General_helper.py

  • qcolor_to_rgba

  • with_alpha

  • blend_colors

  • build_chart_theme

Plotting of isolated values

When occasional values are received interspersed with frequent values from another measurement, the parser and ring buffer correctly keep all channels on the same sample timeline and leave missing entries as NaN. This preserves alignment between slow and fast measurements, but sparse channels then appear as isolated finite values surrounded by NaN. In pyqtgraph, those NaN values break line segments. In fastplotlib, the current compaction of finite values risks connecting points across time gaps. We need a plotting approach that displays sparse event-like values together with dense continuous signals while keeping plotting fast and efficient.

[Status: implemented for pyqtgraph and fastplotlib automatic mode] [Plan:

  • Keep the current parser and circular buffer structure unchanged so all channels remain aligned on the same sample-number axis and sparse channels continue to be represented by NaN gaps.
  • Add a plotting-level render mode concept for traces, with at least line, scatter, and hybrid behavior.
  • Use auto as the default engagement mode and do not add new main-window controls in the first implementation.
  • In auto, classify each visible trace from the current data pattern and choose between line-only, scatter-only, or hybrid behavior.
  • Keep the first rollout UI-free. If manual override is needed later, add it as a context-menu action on the plot or legend rather than consuming permanent screen space.
  • In pyqtgraph, keep the existing line trace with connect='finite' for continuous data and add a scatter overlay for isolated points detected from the visible NaN pattern. Done.
  • Detect isolated values efficiently per channel using finite-neighbor checks, for example a point that is finite while both the previous and next samples are non-finite.
  • Keep pyqtgraph legend and visibility behavior tied to the main line item while the scatter overlay remains an internal helper trace. Done.
  • In fastplotlib, stop treating all finite values as a single compacted line when that would bridge time gaps. Preserve segment breaks by inserting separator rows for discontinuities and render isolated samples with a native scatter overlay. Done.
  • Add a simple heuristic or per-channel mode selection so dense channels remain line plots while sparse channels can automatically switch to scatter or hybrid display.
  • In fastplotlib, rebuild the visible trace buffer from the current window so line continuity follows the original sample timeline instead of the old compacted append-only representation. Done.
  • Keep autoscaling and legend behavior consistent across both backends, including visibility toggles and efficient min/max computation for sparse channels.
  • Implement the pyqtgraph version first because it already preserves NaN segment breaks cleanly, then bring fastplotlib to feature parity with the same visual behavior.]

BLE Issues

Repair checkbox-controlled Serial/BLE routing without breaking line-oriented commands, then determine whether duplicate BLE output remains.

[Status: command-input regression fixed in code; pending live BLE/USB validation] [Findings:

  • In the previous committed design, transport connection state controlled command routing. The Serial and BLE Display checkboxes only controlled terminal rendering, so a command was sent to every connected transport even when its Display checkbox was clear.
  • The first checkbox-routing implementation correctly separated Serial and BLE targets, but its per-transport payload builder omitted the final line terminator. For example, "." became 2E instead of 2E 0D 0A, leaving the MAX30001G line reader waiting for the rest of the command.
  • The first implementation also changed terminal callback connections whenever a Display checkbox changed. This was unnecessary because the Serial and BLE display handlers already consult their display flags, and it introduced avoidable connect/disconnect errors.
  • The reported duplicate BLE command or duplicate terminal output is not yet confirmed as a second defect. Both checked sources can legitimately show the same device output when the device is connected through USB and BLE. Qt receive connections already request UniqueConnection, so rejected repeated connections produce log errors rather than duplicate callbacks.
  • Keep this separate from the MAX30001G/BLESerial large help-output truncation investigation, whose transport-congestion hypothesis remains unconfirmed.] [Implemented:
  • Command entry, multiline paste, and Send File target only checked transports that are currently ready.
  • Each selected transport uses its own line-ending setting.
  • Command payload construction now honors all configured EOL choices exactly: none, CRLF, LF, CR, and LFCR. Focused tests cover single commands, multiline text, trailing newlines, empty input, and every configured EOL.
  • Restored the prior stable terminal callback lifecycle; Display checkboxes filter rendering without reconnecting worker signals.
  • Initialized the main recording state before checkbox callbacks can run, and recording now follows checked Display sources.
  • Reverted the speculative BLE worker start/stop guard; notification lifecycle changes require evidence from live reconnect testing.
  • Updated command-input tooltips to describe checkbox-controlled routing.] [Plan:
  1. Run SerialUI with BLE checked and Serial clear. Confirm . and z each change MAX30001G state once and that the device receives the configured terminator exactly.
  2. Repeat with Serial checked and BLE clear, then with both checked. Confirm one write per checked, connected transport.
  3. Record with one source checked at a time and verify that only that source writes to the file.
  4. Reconnect BLE several times and verify there is one notification subscription and one terminal append per notification.
  5. If duplicate BLE output remains with Serial clear, add temporary sequence logging at the final BLE write and notification boundaries to distinguish duplicate GATT activity from terminal rendering.
  6. Remove temporary diagnostics after the remaining defect is proven.
  • Acceptance criteria
  • . and z work over BLE and USB with the selected line ending.
  • Unchecked or disconnected transports receive no command or file payload.
  • One user action causes one write per checked, connected transport.
  • One BLE notification is displayed once when only BLE Display is checked.
  • Recording contains only checked sources.
  • BLE reconnects do not multiply callbacks or notification subscriptions.
  • Report hardware results separately from compile and unit-test evidence.]