Use date column to calculate stay_len and stay_id

[chenyangkang]

The current add_stay_id function is using number of the rows as stop length, this is valid if the timesteps are sampled in the same frequency.

BirdFlowR/R/route.R

Lines 176 to 184 in cef92ce

	add_stay_id <- function(df) {
	# Benjamin's function
	df |>
	dplyr::mutate(stay_id = cumsum(c(1, as.numeric(diff(.data$i)) != 0)),
	stay_len = rep(rle(.data$stay_id)$lengths,
	times = rle(.data$stay_id)$lengths))
	}
	points <- points |> dplyr::group_by(.data$route_id) |> add_stay_id()

However, custom data (e.g., tracking, motus, banding) seldom contains equally sampled timepoints. So consider using "real" calculation on the date column to get stay_len with default unit of day.

add_stay_id or similar transformation will be a default behavior for BirdFlowRoutes.

I removed the add_stay_id in route function.

BirdFlowR/R/route.R

Line 187 in 83eaf15

# add_stay_id <- function(df) {

And add the function add_stay_id_with_varied_intervals

BirdFlowR/R/RouteDataFunction.R

Lines 395 to 415 in 83eaf15

	add_stay_id_with_varied_intervals <- function(df, timestep_col = "date", timediff_unit = "days", time_threshold = Inf) {
	# Ensure the data is sorted by timestep
	df <- df |> dplyr::arrange(.data[[timestep_col]])
	new_df <- df |>
	dplyr::mutate(
	timestep_diff = c(1, as.numeric(diff(.data[[timestep_col]]), units = timediff_unit)), # Time differences
	i_change = c(1, as.numeric(diff(.data$i)) != 0), # Changes in 'i'
	stay_id = cumsum(i_change | (timestep_diff > time_threshold))
	) |>
	# Now the stay_id is assigned, calculate the duration (time difference) of each stay
	dplyr::group_by(route_id, stay_id) |>
	dplyr::mutate(
	stay_len = as.numeric(max(.data[[timestep_col]]) - min(.data[[timestep_col]]) + 1, units = timediff_unit)
	) |>
	dplyr::select(-timestep_diff, -i_change)
	return(new_df)
	}

which is applied when a new BirdFlowRoutes object is created:

BirdFlowR/R/RouteDataClass.R

Lines 198 to 209 in 83eaf15

	## Add stay id
	birdflow_route_df <- birdflow_route_df |>
	sort_by_id_and_dates() |>
	dplyr::group_by(.data$route_id) |>
	add_stay_id_with_varied_intervals(timestep_col = timestep_col, timediff_unit = timediff_unit) |>
	# Here, using add_stay_id_with_varied_intervals, rather than add_stay_id.
	# It takes 'timestep' as input so account for varying intervals,
	# if the data is not sampled in a frequency.
	dplyr::ungroup() |>
	as.data.frame() |>
	preserve_s3_attributes(original = birdflow_route_df)

Also now the synthetic routes generated by route function will not have circular dates (so cross the year boundary), but the timestep will circulate to 1 again. So we should calculate the stay based on date rather than timestep.

This change will be included in the next merge if nobody objects.

[ethanplunkett]

Make sure you do some testing with plot_routes(). It will likely need some updating for the new format. I've wanted to drop the circular dates and drop a bunch of hackish stuff I did to deal with the circular dates, so this is an overdue change. Let me know if you want me to make the changes in the plotting.