[upstream rENA 0.3.1] Regression axes named after the first edge, not the predictor

[HUDongpin]

This is an issue in rENA 0.3.1, not in ena-python. It is tracked here because it came out of porting rENA to Python. Reported as an observation, not an accusation — it may be fixed upstream or intended, in which case please correct me. Full write-up: docs/rena-upstream-issues.md

Summary

ena.rotate.by.hena.regression names its rotation axes after the first edge rather than the predictor. With both x_var and y_var, both columns get the same name.

Severity: low — cosmetic, but produces duplicate column names.

Setup (rENA 0.3.1, R 4.4.2)

library(rENA)
rows <- data.frame(
  unit  = rep(paste0("U", 1:6), each = 4),
  conv  = rep(rep(c("c1", "c2"), each = 2), times = 6),
  grp   = rep(c("g1", "g2"), each = 12),
  score = rep(c(10, 20, 30, 40, 50, 60), each = 4),
  A = c(1,0,1,0, 0,1,1,1, 1,1,0,1, 0,0,1,0, 1,0,0,1, 0,1,1,0),
  B = c(0,1,1,1, 1,1,0,0, 0,1,1,0, 1,0,0,1, 0,1,1,0, 1,0,0,1),
  C = c(1,1,0,0, 1,0,1,0, 1,0,0,1, 0,1,1,1, 1,1,0,0, 0,0,1,1),
  D = c(0,0,1,1, 0,1,0,1, 0,1,1,0, 1,1,0,0, 0,0,1,1, 1,1,0,0),
  stringsAsFactors = FALSE)
acc <- ena.accumulate.data(
  units = rows[, "unit", drop = FALSE], conversation = rows[, "conv", drop = FALSE],
  metadata = rows[, c("grp", "score")], codes = rows[, c("A","B","C","D")],
  window.size.back = 2)
set <- ena.make.set(acc, dimensions = 2)

Reproduce

colnames(ena.rotate.by.hena.regression(set, list(x_var = "V ~ score"))$rotation)
#> [1] "A & B_reg" "SVD2" "SVD3" "SVD4" "SVD5" "SVD6"
#         ^ expected "score_reg"

colnames(ena.rotate.by.hena.regression(set,
  list(x_var = "V ~ score", y_var = "V ~ grp"))$rotation)
#> [1] "A & B_reg" "A & B_reg" "SVD3" "SVD4" "SVD5" "SVD6"
#         ^ both axes get the same name

Root cause

ena.rotate.by.regression.R:75 derives the name with xName <- all.vars(x)[2], which assumes x is a formula object. params$x_var is a character string, and all.vars() on a string returns nothing:

all.vars("V ~ score")               #> character(0)   -> all.vars(x)[2] is NA
all.vars(as.formula("V ~ score"))   #> [1] "V" "score"

So is.na(all.vars(x)[2]) is always TRUE and the branch falls back to xName <- names(v1)[1] — the first edge name. Line 121 does the same for yName, so with both axes the columns collide.

Suggested fix

Coerce first, e.g. all.vars(as.formula(x))[2].

What ena-python does

Names the axis after the predictor (score_reg, grp_reg). Names are cosmetic and the vectors match, so ena-python does not reproduce this.

[HUDongpin]

Reported upstream: rENA #50 (2026-07-16).

The report there pins its source references to rENA main @ a09a4b20, verified byte-identical to the released 0.3.1 this was found against — so it applies to their current tree, not just the CRAN build.

This issue stays open here as ena-python's record of the divergence; it can close when the upstream one resolves.