Split SIPNET inputs into multiple files based on start and end time
Source:R/split_inputs.SIPNET.R
split_inputs.SIPNET.RdSubset each SIPNET input file and write a new file containing values >= start.time and < end.time (note: end time is not inclusive)
Usage
split_inputs.SIPNET(
start.time,
stop.time,
inputs,
overwrite = FALSE,
outpath = NULL,
time_resolution = c("row", "day"),
allow_varying_timesteps = TRUE
)Arguments
- start.time
Start date or datetime for splitting
- stop.time
End date or datetime for splitting
- inputs
Named
inputslist as provided by PEcAnsettings. Must have structure like:list(met = list(path = "path/to/sipnet.clim"), events = list(path = "path/to/events.in"), ...)- overwrite
if
TRUE, overwrite existing target file (DefaultFALSE)- outpath
if specified, write output to a new directory. Default
NULLwrites back to the directory being read- time_resolution
One of "row", "day". Should the split happen at the closest row to the end time or align with a time unit? See details
- allow_varying_timesteps
logical: Accept files where not all rows have the same duration? Disallowing can be substantially faster.
Value
Modified inputs list with all path entries replaced with new
paths (suitable for inserting back into settings$run$inputs).
Details
NOTE that sipnet met files contain dates and times, while sipnet event files contain only dates. Comparing a datetime to a date will coerce the date to midnight UTC.
Efficiency
Splitting met files goes faster by far when all the inputs are known to have
a uniform timestep, fixed number of characters per line, and all splits are
at day boundaries. Setting time_resolution = "days" and
allow_varying_timesteps = FALSE opts in to a fast code path that finds the
wanted lines of the file by computing byte offsets and therefore avoids
needing to read every row of the input file.
For small subsets of large files this can be several times faster.
If the input has variable line length but constant timesteps (as determined
by checking the first 10 lines; see the n_head argument of
subset_met_lines()), the fast path is taken but is not as fast: it
must scan from the beginning to count line separators, but still skips the
tail of the file after stop.time is reached.