In evanmascitti/soiltestr: Functions for Laboratory Soil Testing

Set object class to "psa" Determine sub-classes based on structure of files in the directory; there are two sub-classes assigned....one for the measurement method and one for the pre-treatment method. This makes the psa class easily extensible with new methods as I develop them! The only thing that must stay constant is the output of the method, and actually this doesn't really NEED to stay constant but it will make it much much easier when analyzing actual data to compare the methods. For example, I can always return the same 4-item list:

• A data frame containing the percent passing for all samples (with arbitrary particle diameters)
• Data frame containing the % falling into each traditionally defined size class
• Data frame containing a plot of the psd
• Data frame or named character vector providing information about the pre-treatment and analysis methods used
• this list could also be extensible to include additional items.....as long as the order is kept constant when new methods are created, any object with class "psa" can still be computed on iteratively without breaking existing methods.

---

2021-03-09

I have decided against using classes and object-oriented programming because it is not really meant for working with data frames. Instead I am going to use the switch() function which still allows this function to be extensible via new helper functions as the PSA methods (in the physical sense) are created.

Some pseudocode:

1. Determine which method was used based on a file inside the directory
2. Calculate the oven-dry specimen mass used in the particle size analysis - this might require a correction due to pretreatements which reduce the initial mass of the specimen
3. Compute the "% finer than x " for an arbitrary number of bins - this is the most complex part of the program. It requires a different method for hydrometer vs pipette, and a way of handling the pipette data for an arbitrary number of input particle sizes (for example, 2 microns only vs 2 and 5 vs 0.2, 2, and 20, etc.). I am sure this can be achieved with across() or something similar, perhaps based on the column positions.
4. Compute the "% finer than x % for all diameters > 53 microns. This too requires a method which can take an arbitrary number of sieves, although this is relatively straightforward. I can use lag().....probably can do that for the above problem too, instead of messing with pivoting wider and then re-pivoting longer again.
5. At this point any following steps will be identical regardless of the method, with the exception of the final list item whose structure will differ based on the method. The contents of the other list items might be different, for example a simple pipette analysis with no sieving will have some NA fields because sub-classes can't be determined. Anyway, at this step, bind the "% finer than" data for the coarse and fine fractions.
6. Calculate other information, like the % of the sample falling into each traditional size class.
7. Build a list to output
8. Return the list

IT WORKS

Later at night, thought of a way to deal with the variable number of pipette samples...start with percent passing tibble, pivot wider across all column names having only digits....then conditionally mutate the table based on the protocol number and switch() to compute the extra size classes only for the protocols that measure them. Actually what would be even better is to make this strucure no matter what but to input NA values if that bin was not measurable.

For now, I am going to only compute the columns gravel, sand, silt and clay as the default action....then if the protocol includes any extra size classes, I can append the table based on the detected protocol ID. This flows with the 'start simple but leave room for extension' philosophy of the psa functions.

evanmascitti/soiltestr documentation built on Oct. 6, 2022, 5:32 p.m.