Believe BINIT only where PSS/E would not have moved it

[luke-kiernan]

Closes #57.

Si is a status, not a step count

A v35 SWITCHED SHUNT block carries a status Si — in service or out — so block i contributes Si * Ni * Bi. number_engaged counts engaged steps, and was being filled with the bare status, understating every multi-step block by a factor of Ni. An in-service block now engages all Ni of its steps; the RAW has no way to say a v35 block is partly in.

BINIT is believed only where PSS/E would not have moved it

BINIT is the device's actual admittance only where PSS/E would not have adjusted the shunt itself:

condition	why
MODSW == 0	locked
bus type 3	PSS/E does not adjust a shunt on the swing bus, whatever MODSW says
MODSW == 2	continuous: the admittance is off the step ladder, so the block statuses could not express it even in principle
no Si column (pre-v35)	nothing to reconstruct from
solved_case = true	caller declares the RAW was written out after a converged power flow

Otherwise — a discretely adjusted shunt (MODSW 1/3/4/5) on a type 1/2 bus — BINIT is only where PSS/E's own solution started, while the engaged blocks are what the record asserts about the device. solved_admittance is then left off entirely, and PowerFlows.jl falls through to number_engaged .* Y_increase; _apply_func! and the OpenAPI reader are both already haskey-guarded, so the absent key needs no other handling.

solved_case (default false) is threaded through parse_file → parse_psse → _pti_to_powermodels!, and is available as a PowerModelsData keyword.

Only v35 cases change behaviour. Pre-v35 records keep BINIT unconditionally and keep their zero-filled number_engaged.

Block reading follows the POM's terminator

_switched_shunt_blocks replaces _is_switched_shunt_block. The v36 DataFormats manual: "Number of steps for block i (1 through 8). The first zero value of Ni or Bi is interpreted as the end of the switched shunt blocks for bus I." The blocks are a contiguous run from block 1 ended by that terminator, so they are now read as one rather than by filtering zeros out of all eight Ni and Bi columns independently.

The two agree on every well-formed record. They differ only on a record with a gap (N2 = 0 alongside a nonzero B3), which the run now reads the way PSS/E does — one block — instead of promoting the columns past the terminator to blocks of their own.

Testing

New fixture test/fixtures/v35_switched_shunt.raw covers each branch: adjusted, locked, continuous, swing-bus, the terminator, and solved_case = true.

774/774 PSS/E parse tests pass (19 new), formatter clean.

⚠️ The OpenAPI half of the suite could not be run. test/Manifest.toml pointed at a deleted path, and re-resolving fails on OpenAPI — PowerCoreOpenAPIModels requires 1.1.x while this branch's compat pins 0.2. That is what #58 fixes; these tests were run against the root project instead. Nothing here touches src/openapi/, so #58 and this PR should not conflict.

Related, not fixed here

PowerFlows.jl's v35 exporter has the mirror image of the Si bug: _build_switched_shunt_steps_v35 writes S_i = engaged[i], pushing a step count into a field PSS/E reads as a 0/1 status. Round-tripping number_engaged = [3, 0] emits S1 = 3.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[Copilot]

🟡 Changes recommended

Moderate findings remain on node-breaker source-bus handling and v35 remote-mode coverage, with one comment nit also outstanding.

Get a fresh assessment by requesting another Copilot review.

Pull request overview

Updates PSS/E v35 switched-shunt parsing for status semantics, BINIT handling, and block termination.

Changes:

• Corrects engaged-step reconstruction and contiguous block parsing.
• Adds conditional BINIT handling and solved_case propagation.
• Adds regression fixtures and parser tests.

File summaries

File	Summary
test/test_parse_psse.jl	Adds v35 switched-shunt regression coverage.
test/fixtures/v35_switched_shunt.raw	Provides cases for BINIT branches, termination, and solved_case.
src/power_models_data.jl	Exposes the solved_case option.
src/pm_io/psse.jl	Implements switched-shunt logic. Findings: moderate (3 votes) source-bus type handling; moderate (1 vote) v35 remote-mode coverage; nit (2 votes) malformed comment.
src/pm_io/common.jl	Threads solved_case through parsing.

Review details

Suppressed comments (1)

src/pm_io/psse.jl:853

• The new v35 behavior depends on all four discrete modes in this tuple, but the fixture only exercises v35 MODSW=1; the existing MODSW=3 test is pre-v35, where has_block_status short-circuits the decision and cannot validate this membership. Please add v35 cases for the remote discrete modes (at least 3, with 4/5 if supported) so a mode-specific regression cannot silently restore BINIT.

const _DISCRETE_SWITCHED_SHUNT_MODES = (1, 3, 4, 5)

• Files reviewed: 5/5 changed files
• Comments generated: 2
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.