Handle ObservationSummary dict is None

[g7gpr]

Avoid iterating over ObservationSummary dict when dict is None

[dvida]

@g7gpr Could you test this latest version with my modifications?

[g7gpr]

Incoporated pr #910, from issue #739, with thanks to @markmac99

Retesting overnight on au0004,6,7

[g7gpr]

This needed a bit more work to catch the edge case where a key that was not a valid column name was not used to create a column, but then tried to write into a column with that name. Also incorporated features from issue #739. However, the changes by @dvida are good.

[g7gpr]

Tested and working on three stations. Ready for further review / merge.

[dvida]

Reviewed this — one correctness regression worth flagging, which I've pushed a fix for (98c1eb8).

storeDictInDB silently drops stationID (and any mixed-case key).

dict_filtered_by_columns = {k: v for k, v in d.items() if k in existing_columns}

Columns are always created lower-cased (addRequiredColumns → ADD COLUMN {key.lower()}), so existing_columns only ever contains lower-case names. The new filter compares the original-case key, so stationID (added at addObsParam(d, "stationID", ...)) is filtered out and never inserted — the stationid column ends up always NULL. Pre-PR code inserted d.items() directly and relied on SQLite's case-insensitive column matching, so it stored fine. This is a data-loss regression for the station identifier.

Fix matches case-insensitively and keys the filtered dict by the lower-cased name:

dict_filtered_by_columns = {k.lower(): v for k, v in d.items() if k.lower() in existing_columns}

Verified in isolation: old filter drops stationID, new one keeps it.

Also ungated the dropped-keys log.warning (was behind debug), so genuine data loss is visible in normal logs rather than silent.

Everything else checked out: the timestampFromNTP 3-tuple change is consistent with its only caller, the None-dict guards are sound, and the non-dict-JSON rejection is a good hardening.

[g7gpr]

Agree with these changes, penultimate commit introduced an error, which latest commit has corrected.

(vRMS) au0004@birch:~/source/RMS$ sqlite3 /srv/rms/RMS_data/AU0004/observation.db -header -column "select commit_date, stationid from observations order by commit_date desc limit 4;"
commit_date      stationid
---------------  ---------
20260629_151259  AU0004   
20260628_020646           
20260628_020646           
20260627_031825  AU0004   

I'm testing on au0004,6,7; there should be enough night left to capture a small amount of data.

[g7gpr]

All looks good now.