Ben
You're right, it's just tire kicking. No special meaning to the values chosen here. In replacing `lagrangian_value` with `primal_value`, I'd suggest ensuring that each subproblem's `primal_value` is no smaller than...
The use case I had in mind is the one described in [comments to #190](https://github.com/markmfredrickson/optmatch/issues/190#issuecomment-696813855): > 1. User fits `psmA` using data `dat`; then `distA = match_on(psmA, data=dat)`. At this...
As things currently stand [proj1-nodeprices ce38fbc], there are a few conditions where we don't record subproblem info, e.g. [solve_reg_fm_prob.R#L87-L94](https://github.com/markmfredrickson/optmatch/blob/2c3ab994f898bedc301988346ed9d6b059ed2511/R/solve_reg_fm_prob.R#L87-L94): ```r if (floor(min.cpt) > ceiling(max.cpt) | #inconsistent max/min ceiling(1/min.cpt) < floor(1/max.cpt)...
@adamrauh reports via email that subproblems tables are showing `feasible=FALSE` for subproblems that are actually feasible: ```r data(nuclearplants) mm
In this case, the original f.m. problem was infeasible but the two f.m. problems that you call out, producing `tmp2` and feeding a call to `return()`, are both feasible. As...
Nice! Could you also add tests asserting of a known infeasible subproblem that it's recorded as such, and also of a known feasible problem that it's recorded as such? And...
The following two lines of `fmatch()` lead me to expect that infeasibility will in these cases manifest itself with `temp$solution` all equal to `0L`, with no `1L`s. With the LEMON...
... in addition, `fmatch()` itself should be using either the appropriate one of these feasibility indicators to populate the feasibility entry in its subproblems table, but instead I see: https://github.com/markmfredrickson/optmatch/blob/4d9e3423adc095e591f3d5824d35c0e2c676d6e1/R/fmatch.R#L319-L325...
With the recent updates for #129, we may be catching a somewhat greater share of these cases than previously, by more reliably flagging subclasses in which the number of controls...
314eee7 addresses the first part and at least some of the second. Remains to do testing.