Useability Improvements

[trhallam]

Just suggestions for consideration.

It would be good if the header information can be attached to the deviation as attributes to contain all the deviation data in a single struct.

Resampling in the TVDSS domain would be a great addition.

loc_to_tvdss changes the depth attribute but there is not easy way to tell which domain the position is then in. Also, if you use resample it still refers to the MD deviation.

A lot of software will only ever store md, inc, azi and all other values are calculated on the fly. If the header information is included as an attribute in the struct then tvdss x, and y can all be live @property methods of the struct.

[Zabamund]

Thank you for your suggestions @trhallam , it's great to get feedback like this.

[...] header information [...]

I'm looking into this, the header is not mandatory so far.

Resampling in the TVDSS domain [...]

currently .resample() is only available on the position object returned by the minimum_curvature method, this is already in the TVDSS domain, could you clarify? Perhaps our docs are not clear enough to illustrate this but this is what we currently have:

step = 30
depths = list(range(0, int(dev.md[-1]) + 1, step))
pos = dev.minimum_curvature().resample(depths = depths)
dev2 = pos.deviation()

dev is always in md, inc, azi and pos is always in depth (TVD or TVDSS), northing, easting

loc_to_tvdss changes the depth attribute [...]

This has actually been renamed to to_tvdss in PR #48. This function doesn't change the depth domain as it's only available on a position object, it will Shift tvd to tvdss given datum elevation. You're right that the change is not "recorded".

[...] if you use resample it still refers to the MD deviation.

could you show me an example of what you mean here?

A lot of software will only ever store md, inc, azi [...]

We have chosen to keep the deviation and position objects seperate and not to assume any particular method is used for conversion.

[trhallam]

This is what I get when I try to use resample on a deviated well. I'm expecting the depths to be whole ints but I get back something different. As per your example, this is because the resampling is done in the MD domain. The confusion here comes that you have a position object which can be resampled, but on the deviation depths not the position depths. Or at least that is how it appears to me. It might be more clear if the key word argument depths was changed to measured_depths or just md.

For seismic at least, it would be more helpful to have constant vertical sampling rather than constant along well-bore sampling.

import wellpathpy as wpp
import numpy as np
import pandas as pd

f12_dev = pd.read_csv("data/well_f12_deviation.asc", comment="#", delim_whitespace=True)
f12_dev_pos = wpp.deviation(*f12_dev[["MD", "INCL", "AZIM_GN"]].values.T)
pos = f12_dev_pos.minimum_curvature()
pos2 = pos.resample(depths=np.linspace(0, 3110, 3111, dtype=int))
pos2.depth[-10:]

array([2858.55757458, 2859.14567965, 2859.73384144, 2860.32205983,
       2860.91033475, 2861.49866611, 2862.08705382, 2862.67549779,
       2863.26399793, 2863.85255414])

[trhallam]

For the other points, my initial hunch is to create a class which abstracts much of the wellpathpy mechanics and takes care of most of the details for a user. The class would contain the deviation and header attributes as well as the preferred method of calculating deviation (which could be set at initialisation time).

Then position could be returned as properties on the fly, accounting for what attributes were available in the class.

class WellBoreSurvey:
    
    def __init__(self, md, inc, azi, method='minimum_curvature', x0=None, y0=None, z0=None):
        
        self._dev = wpp.deviation(md, inc, azi)
        self.method = method
        self.x0 = x0
        self.y0 = y0
        self.z0 = z0
        
    @property
    def position(self):
        pos = getattr(self._dev, self.method)()
        if self.x0 is not None and self.y0 is not None:
            pos.to_wellhead(self.x0, self.y0, inplace=True)
        return pos
    
    @property
    def position_tvdss(self):
        pos = self.position
        if self.z0 is not None:
            pos.loc_to_tvdss(self.z0, inplace=True)
        return pos
        
        
wbs = WellBoreSurvey(*f12_dev[["MD", "INCL", "AZIM_GN"]].values.T, x0=6478566.23, y0=435050.21, z0 =54.9,)
print(wbs.position.depth[:10])
print(wbs.position_tvdss.depth[:10])

[  0.         145.9        159.94999743 173.99997946 188.99995178
 203.99993325 218.99991415 233.99987665 248.99980779 263.99969323]
[  54.9         -91.         -105.04999743 -119.09997946 -134.09995178
 -149.09993325 -164.09991415 -179.09987665 -194.09980779 -209.09969323]

Of course, you could also include some default sample rates for output and helpers etc, like a to_dict method to make building dataframes super easy.