lpspec holds a queue of language questions for this repository, and they arrive in families #231

Activity

1. FBumann

   added

   metaAbout the project rather than the language — trackers, naming, processAbout the project rather than the language — trackers, naming, process

   on Aug 31, 2026
2. FBumann commented on Aug 31, 2026

   

   FBumann

   on Aug 31, 2026

   Last edited by FBumann

   ContributorAuthor

   More actions

   Prompt: Maybe we should transfer all of them from lpspec to mathspec and close them in lpspec? Lets go through one by one.

   Note

   The following content was generated by AI.

   The label is empty, so this closes. Triaged one by one; twelve language questions now live here, seven were answered or declined without one, and two stayed upstream as lpspec work.

   Moved here — each closed upstream, its body left readable as the evidence these link to rather than copy:

   lpspec	here
   #851	#255	a where-comparison may name a co-located column — decided yes upstream
   #982	#258	may a where: carry operators — relocation, reduction
   #218	#259	a reduction cannot filter the terms it sums
   #1219	#260	N declarations that are one requirement — the residue #168 does not reach
   #31	#261	a bound cannot be an expression
   #1220	#262	bounds: has no where: (language half only)
   #383	#263	domain: semi_continuous (language half only)
   #384	#264	cumsum over a variable-free expression (language half only)
   #220	#265	an indicator constraint has no spelling
   #1113	#266	a binary times a bounded variable
   #261	#267	a declared McCormick relaxation (language half only)
   #1222	#268	a data combination that cannot be built has no load-time refusal

   Answered or declined, with no issue here:

   • docs: a page says how the package version and the language version of a model file relate #711, #1129 — answered by Nested constraints and expressions #2, implemented in feat: a named expression may give a value per region, and no two regions may claim one coordinate #168: expressions: takes cases: with an otherwise:, no two when masks able to hold at once, proved at load.
   • feat(language): a where may compare arithmetic over parameters #469 — subsumed by may a where: carry operators — relocation through a lookup, reduction over a dimension it does not span #258, which carries relocation and reduction as one decision.
   • #1014 — no witness whose boundary is neither first nor last, its own close condition; and downstream of may a where: carry operators — relocation through a lookup, reduction over a dimension it does not span #258 either way.
   • docs(pypsa): a risk preference with weight zero builds the CVaR variables and refuses quadratic costs, as in PyPSA #849 — track spent: two children shipped, one decided, one moved.
   • test: generic spine data and coverage gates, so a silent regime cannot pass — and two wrong formulas they caught #125 — unit checking declined along with conversion. Wants a row in ceiling.md's deliberate non-primitives table making the distinction explicit; that is the one follow-up this sweep leaves in the tree.

   Stayed upstream, delabelled: #1220 rescoped to the port that decides whether its language half is worth anything, and #229, whose remaining variant is a post-bind check in lpspec's build path.

   The policy the sweep settled, for the constructs with a two-sided implementation: the language question is filed here and the lpspec issue closes now, with an adoption issue refiled there when the construct ships. One issue, one PR, on both sides — a combined adoption tracker was rejected for bundling separable work behind one number.
3. FBumann

   closed this as completedon Aug 31, 2026