Some workout columns should be float, not text

[simonw]

Columns duration, totalDistance and totalEnergyBurned should be converted to float.

healthkit-to-sqlite/healthkit_to_sqlite/utils.py

Lines 50 to 57 in 71e36e1

	def workout_to_db(workout, db, zipfile=None):
	record = dict(workout.attrib)
	# add metadata entry items as extra keys
	for el in workout.findall("MetadataEntry"):
	record["metadata_" + el.attrib["key"]] = el.attrib["value"]
	# Dump any WorkoutEvent in a nested list for the moment
	record["workout_events"] = [el.attrib for el in workout.findall("WorkoutEvent")]
	pk = db["workouts"].insert(record, alter=True, hash_id="id").last_pk

[Mjboothaus]

Hi Simon -- just experimenting with your excellent software!

Up to this point in time I have been using the (paid) HealthFit App to export my workouts from my Apple Watch, one walk at the time into either .GPX or .FIT format and then using another library to suck it into Python and eventually here to my "Emmaus Walking" app:

https://share.streamlit.io/mjboothaus/emmaus_walking/emmaus_walking/app.py

I just used healthkit-to-sqlite to convert my export.zip file and it all "just worked".

I did notice the issue with various numeric fields being stored in the SQLite db as TEXT for now and just thought I'd flag it - but you're already self-reported this issue.

Keep up the great work!

I was curious if you have any thoughts about periodically exporting "export.zip" and how to just update the SQLite file instead of re-creating it each time. Hopefully Apple will give some thought to managing this data in a more sensible fashion as it grows over time. Ideally one could pull it from iCloud (where it is allegedly being backed up).

[Mjboothaus]

P.s. wondering if you have explored using the spatialite functionality with the location data in workouts?

[simonw]

Unfortunately I don't think updating the database is practical, because the export doesn't include unique identifiers which can be used to update existing records and create new ones. Recreating from scratch works around that limitation.

I've not explored workouts with SpatiaLite but that's a really good idea.

[Mjboothaus]

In terms of unique identifiers - could you use values stored in HKMetadataKeySyncIdentifier?