Changelog

Transition from Travis to GitHub actions for continuous integration testing.
No new tests.
- Still at 80 tests, all passing.
- Test coverage remains at 100 %.

Compatibility with removal of element prefixes to function names in matsbyname.

Fixed a bug in the website. Vignette wasn’t showing up due to capitalization error.

Now including pkgdown website.
Removed _colname from argument names.
When calling matsindf::collapse_to_matrices, now using matvals argument name instead of values.

Now calculating the D vector by D_j = exp(deltaV_j / L(V)). Doing so reduces the possibility that we’ll hit an overflow error. Previous approach was D_j = exp(deltaV_j)^(1/L(V)). The two approaches are equivalent, mathematically. But the new approach avoids taking exp(deltaV_j), which could overflow when deltaV_j was greater than about 725.

lmdi now defaults to supplying very small values for missing categories (1e-10) when calculating Z matrices.
Both lmdi and Z_byname functions now accept fillrow arguments. fillrow provides an alternative to assuming small values for missing categories. The fillrow approach is especially useful when a row in X_0 or X_T is missing compared to the other, due to a particular category being present in one year but absent in the other. In this situation, we want to fill values in the missing row with non-zero numbers for all categories except for allocation to the final subsubcategory. Then, we set only the allocation from subcategory to subsubcategory (phi_ij) to 0. This approach correctly models the fact that despite the fact that we have no useful exergy of this type being produced in one of the years, we still have total primary en/xergy (E.ktoe), there is still an allocation of primary exergy to the subcategory (phi_i), and if there were machines making this subsubcategory of useful exergy in this time period, they would have a certain primary-to-useful efficiency (eta_ij). It turns out that we don’t need to know the exact values of primary exergy (E.ktoe), allocation to subcategory (phi_i), or primary-to-useful efficiency (eta_ij). These values must simply be non-zero so long as allocation from subcategory to subsubcategory (phi_ij) is zero.

lmdi() now creates first row with 0s (for ΔV terms) and 1s (for D terms). This change means that the outgoing data frame has same number of rows as the incoming data frame, eliminating the need for head or tail padding.

Initial version.

LMDIR 0.1.13 (2020-04-12)