Use PowerIO 0.9 for case and GOC3 parsing

.github/workflows/ci.yml

@@ -30,9 +30,9 @@ jobs:
       fail-fast: false
       matrix:
         julia-version: ['1']
-        # Each backend, and the GOC3 smoke test, run as their own job so that the wall
-        # clock is the slowest slice rather than the sum of all of them.
-        selection: ['nothing', 'cpu', 'cuda', 'goc3']
+        # Each backend, GOC3, and AOT run as their own job so that the wall clock is the
+        # slowest slice rather than the sum of all of them.
+        selection: ['nothing', 'cpu', 'cuda', 'goc3', 'aot']
     steps:
       - name: Checkout repository
         uses: actions/checkout@v4

Project.toml

@@ -3,10 +3,10 @@ uuid = "2fff4b78-0b6c-428d-bac8-85ccea8c4bdf"
 version = "0.4.0"
 
 [deps]
-DelimitedFiles = "8bb1440f-4735-579b-a4ab-409b98df4dab"
 ExaModels = "1037b233-b668-4ce9-9b63-f9f681f55dd2"
 ExaPowerIO = "14903efe-9500-4d7f-a589-7ab7e15da6de"
 JSON = "682c06a0-de6a-54ab-a142-c8b1cf79cde6"
+PowerIO = "05ed8b54-f668-4096-9d0d-e8c3dd9dc169"
 
 [weakdeps]
 ExaModelsCompiler = "3d1e9a26-5b74-4f0c-9a2b-7c8f4e11d3a7"
@@ -18,6 +18,7 @@ ExaModelsPowerExaModelsCompilerExt = "ExaModelsCompiler"
 ExaModels = "0.11"
 ExaPowerIO = "0.3"
 JSON = "0.21"
+PowerIO = "0.9"
 julia = "1.11"
 
 [sources]

README.md

@@ -7,10 +7,26 @@ ExaModelsPower.jl is an optimal power flow models using ExaModels.jl
 [![doc](https://img.shields.io/badge/docs-dev-blue.svg)](https://madsuite.org/ExaModelsPower.jl/dev/) 
 [![codecov](https://codecov.io/gh/MadNLP/ExaModelsPower.jl/graph/badge.svg?token=ybOObxcXhB)](https://codecov.io/gh/MadNLP/ExaModelsPower.jl)
 
+## Case files
+
+Every model constructor takes a path. Case files are read through
+[PowerIO](https://github.com/eigenergy/PowerIO.jl), so MATPOWER (`.m`), PSS/E raw
+(`.raw`), PowerWorld aux (`.aux`), PowerModels JSON, PSLF (`.epc`), pandapower JSON,
+egret and GOC3 JSON all work. The format is inferred from the extension; pass `from` when
+it cannot be, as for a `.json` file:
+
+```julia
+model, vars, cons = ac_opf_model("case118.json"; from = "powermodels")
+```
+
+A bare name that is not a file in the working directory is looked up in the bundled
+PGLib-OPF library, which is what `"pglib_opf_case118_ieee.m"` below resolves through.
+Paths with a directory component are passed to PowerIO unchanged.
+
 ## Usage
 ### Static optimal power flow
 ```julia
-using ExaModelsPower, MadNLP, MadNLPGPU, CUDA, ExaModels, GOC3Benchmark, JSON
+using ExaModelsPower, MadNLP, MadNLPGPU, CUDA, ExaModels, JSON
 
 
 model, vars, cons = ac_opf_model(
@@ -60,6 +76,12 @@ result = madnlp(model; tol=1e-4)
 ```
 
 ### Multi-period optimal power flow
+
+The `.Pd` / `.Qd` files hold **MW**, one row per bus in the case's own bus order and one
+column per period; they are converted to per unit on the case's base MVA. To pass the same
+matrices in memory rather than on disk, give both `pd` and `qd` as keywords — one without
+the other is an error rather than a silent half-applied load.
+
 ```julia
 model, vars, cons = mpopf_model(
     "pglib_opf_case118_ieee.m", # static network data

src/ExaModelsPower.jl

@@ -1,15 +1,14 @@
 module ExaModelsPower
 
 import ExaModels: ExaModels, ExaCore, @add_var, @add_par, @add_con, @add_obj, @add_con!, ExaModel, convert_array, solution
-using DelimitedFiles
-using ExaPowerIO
+using ExaPowerIO            # the bundled PGLib-OPF directory, and nothing else
 using JSON
+using PowerIO
 
 # Parsers and data stay separate from the models that consume them.
-include("parser.jl")          # matpower -> the tables every model is built from
+include("parser.jl")          # case file -> the tables every model is built from
 include("constraint.jl")      # the algebraic expressions, shared by all three
-include("goc3_parser.jl")     # GOC3-specific parsing
-include("sc_parser.jl")       #   "
+include("sc_parser.jl")       # GOC3-specific parsing
 
 include("opf.jl")             # static: polar, rect, DC
 include("mpopf.jl")           # multi-period: polar, rect, DC

src/goc3.jl

@@ -10,12 +10,10 @@ function goc3_model(
 
 
     uc_data = JSON.parsefile(uc_filename)
-    data = get_data_from_file(filename)
-    data_json = JSON.parsefile(filename)
-    sc_data, lengths, producers_first = parse_sc_data(data, uc_data, data_json)
+    sc_data, lengths = parse_sc_data(read(filename, String), uc_data)
     @info "parsed data"
 
-    (L_J_xf, L_J_ln, L_J_ac, L_J_dc, L_J_br, L_J_cs,
+    (; L_J_xf, L_J_ln, L_J_ac, L_J_dc, L_J_br, L_J_cs,
     L_J_pr, L_J_cspr, L_J_sh, I, L_T, L_N_p, L_N_q, L_W_en_min_pr,
      L_W_en_min_cs, L_W_en_max_pr, L_W_en_max_cs, K) = lengths