Setting data in `ts` and `ui`, pass by reference in python functions

[rburghol]

Overview

The most basic foundation of HSPF Special Actions was the creation of custom state variables, and updating of system state variables. Determine how to do this in hsp2

• [x] How does Python/hsp2use the h5 file data?
  • Does it set values back to the timeseries in the h5 file?
  • Are there implicit pass by reference variables, for example in main _hydr_ loop initialization (before the actual timestep loop):
  • ts['ROVOL'] = ROVOL = zeros(steps)
  • Initializes the ROVOL in ts, also initializes a local variable ROVOL as a complete array for every step of the model run
  • Then, ROVOL[step] = rovol / VFACT is the only place in the code where any value storing occurs, suggesting that ts['ROVOL'] is aliased to ROVOL, and therefore, the storage is easy.
  • From https://github.com/respec/HSPsquared/blob/8176be9236be9ee6e55422d01531591a769a8df7/HSP2/HYDR.py#L209
• [x] Can we pass inputs by reference, for example:
  • Facility "A" uses flow in "Reach 001" to determine available water, i.e.: available_mgd = (Qreach - flowby) / 1.547
  • "Reach 001" flow is in /timeseries/rchres_001/rovol
  • can we just say (in the class code):
  • self.state['Qreach'] = ds['/timeseries/rchres_001/rovol'] and have it stay updated automatically? No need to call "getInputs()" method?
• [x] Are there other use cases for the "getInputs()" method?

Writing Referenced ts values with sub-components

Work through ts writing. See also:

• Standard Way:
  • Write using save_timeseries
  • called in main.py: https://github.com/HARPgroup/HSPsquared/blob/f9555f2e48cc3409f33c29d3c6504d9bdc0f91fb/HSP2/main.py#L218
  • Function https://github.com/HARPgroup/HSPsquared/blob/f9555f2e48cc3409f33c29d3c6504d9bdc0f91fb/HSP2/utilities.py#L288
• For SPECL path must be more explicit, because default method is to set ts[varname] where varname can cause name collisions:
  • specific BLOCK (low risk): low risk because individual variables in HSPF are quite distinct, with very little overlap between blocks (tho this might be possible)
  • segment ID: very high risk of overwrite, as for example one may read or apply special actions to multiple rchres or perland segments, and there is no way to discern them.
• i.e. desire to read vars as:
  • Qin = ts[/RESULTS/RCHRES_001/IVOL] + ts[/RESULTS/RCHRES_002/IVOL]
• and write vars as:
  • save_timeseries(store, ts, ui_specl['SAVE'], siminfo, saveall, 'SPECL', null, null)
  • with save_timeseries knowing that if it gets the name "SPECL", or perhaps a more generic argument value that indicates "full path", it will iterate through the keys and parse them into the path constituents by splitting on "/"
  • OR, just use direct hdf5 write routine data_frame.to_hdf(self._store, path, format='t', data_columns=True, complevel=complevel)

    
    import numpy as np
    from numpy import zeros, any, full, nan, array, int64
    steps  = 5# adapted from HDRY:steps = int(ui['steps'])
    ts['/RESULTS/RCHRES_001/SPECL/Qin'] = zeros[steps]

initialize the columns... very important

ts = { '/RESULTS/RCHRES_001/SPECL/Qlocal':zeros(steps), '/RESULTS/RCHRES_001/SPECL/Qin':zeros(steps), '/RESULTS/RCHRES_001/SPECL/Qout':zeros(steps) }

now populate some values into the Qlocal variable

ts['/RESULTS/RCHRES_001/SPECL/Qlocal'][3] = 2.0 ts['/RESULTS/RCHRES_001/SPECL/Qlocal'][1] = 7.0

do this calc in direct evaluation

ts['/RESULTS/RCHRES_001/SPECL/Qout'] = ts['/RESULTS/RCHRES_001/SPECL/Qlocal'] + ts['/RESULTS/RCHRES_001/SPECL/Qin']

do calculation using indirect evaluation via eval()

ts['/RESULTS/RCHRES_001/SPECL/Qout'] = eval(\"ts['/RESULTS/RCHRES_001/SPECL/Qlocal'] + ts['/RESULTS/RCHRES_001/SPECL/Qin']\")

print out the result

ts['/RESULTS/RCHRES_001/SPECL/Qout'] array([0., 7., 0., 2., 0.])