Skip to content

Saving and exporting

The Export stage saves your session and writes files for use elsewhere. Everything the web logger holds — time series, spectra, transfer functions, sonograms, modal fits, calibration and units — can be saved to pydvma's native format or exported to MATLAB, CSV or a figure.

Save the session (.dvma)

Save Dataset writes a .dvma container: a zip of a JSON manifest plus plain .npy arrays, containing no executable code and safe to share. It captures the full dataset, including your calibration, units, channel labels and any modal fit — so reopening it (in the web logger, in Python via dvma.load_data(), or in the JupyterLite notebook) restores exactly where you were.

What Save writes: data with its processing

Saving stores the results you computed, not just the raw captures. Any FFT and transfer function on screen is written into the file as a real FreqData / TfData item (coherence included), linked to the measurement it came from. Reopening the file draws those views immediately — no engine, no waiting, and the same items load in Python:

data = dvma.load_data('session.dvma')
data.freq_data_list      # the spectra you computed in the app
data.tf_data_list        # transfer functions, with coherence

Each stored result also records how it was made: the analysis settings in force (window, averaging, channel choice), and a short signature of the source samples it was computed from.

Re-saving updates the results in place — one FFT and one TF per measurement, however many times you save.

"⚠ source changed"

The signature is what lets the app tell an intact result from a stale one. If the time data changes after a result was stored — you resampled it, cleaned an impulse, or edited the samples in a notebook and pushed them back — the stored spectrum no longer belongs to the samples sitting beside it. On load, that measurement's tray card shows a ⚠ source changed badge; click it to recompute the affected views (the badge clears), or leave it and the file keeps what it always had. A file saved by an older pydvma carries no signature and is never flagged — absence of a signature is not evidence of staleness.

Sonograms are asked for

A sonogram is not stored automatically. The view is one channel's picture; the file wants the full complex sonogram, so storing it means running the transform again — slower to save and a bigger file. If you computed a sonogram this session, Save asks Include sonogram data?:

  • This channel — store the channel you are looking at (the default);
  • All channels — store every channel of that measurement;
  • Don't include — skip it. Sonograms already in the document stay there; this only declines the new one.

Never opened the Sonogram stage? You will never see the dialog. Answer it once and an unchanged session will not ask again on the next save.

Not everything is materialised yet

Power spectral densities and cross-spectra are not written as stored results — recompute them after loading. Neither is an ensemble ("across sets") transfer function: it is derived from several measurements at once, and a result that named only one of them as its source could not honestly be checked for staleness. Both are on the follow-up list.

Choose which measurements to save

Each of Save Dataset, Export Matlab and Export CSV is a split control: the button itself always means everything, and the ▾ beside it opens Choose sets…, a tick list of your measurements (with badges showing what each carries — time · fft · tf · fit).

A subset save writes the chosen measurements and everything hanging off them — their spectra, transfer functions and any modal fit that spans them — and nothing else. The pick applies to that one save: it starts all-ticked every time, is never remembered, and is deliberately unrelated to what is shown, faded or hidden on the plot. A subset save also does not clear the autosave, since it is not the whole session.

One thing a subset save does not narrow: results it materialises join your live session as well as the file. Say yes to the sonogram prompt during a subset save and that sonogram is now part of the session too — it will be in the next full save, and it is what a restore or session.data brings back. Only the file is filtered.

The same split exists in Python:

# one measurement plus everything derived from it
dvma.save_data(dataset, filename='just-4.dvma', sets=[3])
small = dataset.subset([0, 3])       # or take the subset in memory

Autosave and session restore

The Autosave switch (on by default) writes your session to browser storage a couple of seconds after every change:

  • If the app has access to a working folder, it keeps an autosave.dvma there.
  • Otherwise it stores the session in the browser (IndexedDB) and offers Restore last session? the next time you open the app.

A clean Save Dataset clears the autosave. Turn Autosave off to stop the background writes.

When the app is served by pydvma-serve, each autosave is also posted to the server, which keeps the authoritative session — so the restore offer on the next open comes from there rather than from browser storage, and a session survives the tab closing or the serve process crashing. See the session journal.

Export data

  • Export Matlab writes a .mat file. The MATLAB bytes are built by SciPy (scipy.io.savemat) in the engine, so the structure matches the Python export_to_matlab.
  • Export CSV writes CSV files (one per data kind — time / freq / tf). The CSV is generated to byte-for-byte match pydvma's export_to_csv: %.18e formatting, complex values written as (RE±IMj), and raw (uncalibrated) values, so a browser export and a Python export of the same data are identical.

Schema parity

Both the .mat and CSV exporters reproduce the Python file schemas exactly, so files are interchangeable between the web logger and scripted pydvma workflows. The .dvma format is the same on both sides too — see the format reference.

Export figures

The current plot can be written as a figure:

  • format — tick PNG (raster, 3× scale) and/or PDF (vector); tick both to write both.
  • background — white (default), transparent, or dark. The dark option recolours the figure chrome for a dark background while preserving the data lines — useful for slides.
  • filename — a default like pydvma_figure_YYYY-MM-DD_HHMM, editable.

The figure contains what the plot shows: the legend is included when it is toggled visible (at its on-screen position, listing the drawn lines — hidden lines are left out), and the TF coherence overlay and its right-hand axis export exactly when that toggle is on. A Bode export contains both stacked panes; a sonogram export includes the heat map.

Press Export to write the ticked formats.

'Dark' here is a figure option, not a theme

The dark background applies only to exported figures and is independent of the app's own light/dark theme toggle — exports are theme-invariant, so the same figure comes out whichever theme the interface is using.

Where files go

If you granted the app a working folder, saves and exports land there; otherwise they download through the browser. Either way the files are ordinary .dvma / .mat / .csv / .png / .pdf you can move, share or reopen.

Opening files

Load data from the header's Load Data button. The web logger opens:

  • .dvma — read directly (no engine needed);
  • legacy .npy pickle files from pydvma ≤ 1.4.0 — decoded by the pyodide engine; and
  • .mat files from the original JW logger — spectral files (spectra / transfer functions, with coherence columns recognised automatically) and time captures.

Format is detected from the file's content, not its extension. See The .dvma file format and From the Qt logger.

Loading adds; it does not replace. With data already present, loading another file appends its sets alongside the current ones — the tray and legend show everything together (the old logger's "Add on load"), and Save Dataset writes the composite to one .dvma. To drop a set, use its tray ×; to start from scratch, reload the page. One caveat: a fitted modal model inside an appended file is ignored — the session keeps its own live fit.