R1-03: Core strong types

[aravgarg28]

R1-03: Core strong types — Completion Report

Status: complete
Branch/PR: task/R1-03-core-types

What was built

src/core/include/microsim/core/types.hpp (+ types.cpp): the dimensioned value types and identifiers every later module depends on, encoding docs/numerics/NUMERIC_REPRESENTATION.md in the type system.

• Value types: Price (int64 ticks), Qty (int64 lots), Cash (int64 minor units), SimTime/Duration (int64 ns), and Side (+ opposite, sign_of, to_cstr).
• Identifiers: OrderId, TradeId, Seq (sequential, with first()/next() for the monotonic assigners), ClientOrderId, ParticipantId, InstrumentId — opaque handles that order and hash but carry no arithmetic.
• Narrow operator sets: Price ± int → Price, Price − Price → int64 ticks, Qty ± Qty, Cash signed arithmetic, SimTime ± Duration → SimTime, SimTime − SimTime → Duration. Meaningless operations simply have no overload, so they are compile errors.
• No implicit int conversions (explicit ctors, named accessors); std::format + ostream printing via integer scale math with unit tags (1003t, 50lot, -25mu, 500ns, order#42). Currency formatting (needs a tick size) is correctly deferred to R1-06 per the task's Excludes.

Acceptance criteria

• All listed types present with the exact NUMERIC_REPRESENTATION §strong-types rules — pinned by static_assert (allowed-op result types + no-implicit-conversion) in test_types.cpp
• static_assert suite for type rules — 14 static_asserts (result types, convertibility, size/triviality)
• std::format/formatter printing via integer scale math — Format.* tests
• Excludes respected — no Notional (needs instrument, R1-05), no string conversions (R1-06)
• Unit tests: ops tables — Price/Qty/Cash/Time/Side/Ids/Format suites
• Compile-fail tests for banned ops — 9 cases via try-style CTest targets (WILL_FAIL), plus a positive control

Verification output

$ cmake --preset debug && cmake --build --preset debug && ctest --preset debug
100% tests passed out of 30
Total Test time (real) = 6.62 sec

$ ctest --preset debug -R core.compile_fail
core.compile_fail.price_plus_qty ........... Passed   (banned op rejected)
core.compile_fail.price_plus_price ......... Passed
core.compile_fail.price_times_qty .......... Passed
core.compile_fail.implicit_int_to_price .... Passed
core.compile_fail.implicit_price_to_int .... Passed
core.compile_fail.qty_plus_cash ............ Passed
core.compile_fail.simtime_plus_simtime ..... Passed
core.compile_fail.qty_times_qty ............ Passed
core.compile_fail.cross_compare_cash_qty ... Passed
core.compile_fail.ok_control_runs .......... Passed   (positive control compiles+runs)

$ cmake --preset asan-ubsan && cmake --build --preset asan-ubsan && ctest --preset asan-ubsan
100% tests passed out of 30                  # no ASan/UBSan findings

$ (build core under -DMICROSIM_WERROR=ON)    # no warnings
$ ./scripts/format.sh --check                # clang-format 20.1.7: 44 files OK

Tests added

Test	Covers	Type
static_assert ×14 (test_types.cpp)	allowed-op result types, no-implicit-conversion, size/triviality	compile-time unit
Price.*, Qty.*, Cash.*, Time.*, Side.*	arithmetic, ordering, sign/opposite	unit
Ids.SequentialIdsIncreaseMonotonically, Ids.HashableAsMapKeys	INV-9 ordering, map-key hashing	unit
Format.StdFormatUsesUnitTags, Format.OstreamMatchesFormat	formatter/ostream output	unit
core.compile_fail.* ×9 + ok_control	banned ops do not compile; positive control does	compile-fail

Deviations / Questions

1. Formatter output is unit-tagged raw integers, not currency. Human-readable currency needs the instrument's tick size, which is R1-05/R1-06 (the task's own Excludes). Price prints 1003t, Cash prints -25mu. This matches "printing via integer scale math" at the level available here; flagging so it's a conscious choice, not an oversight.
2. Two small affordances beyond the literal type list, both spec-justified: Duration * int64 (latency composition needs base * n, LATENCY_MODEL.md) and sign_of(Side) (fill→position sign, used pervasively by accounting). Neither weakens the banned-op guarantees. Happy to remove if you'd rather keep the surface minimal.

Notes for reviewer

• Start at the class comment blocks in types.hpp — each states its allowed/banned operations in prose next to the code.
• The compile-fail harness (tests/CMakeLists.txt, microsim_add_compile_fail) is the interesting infra: each banned op is an EXCLUDE_FROM_ALL target a CTest builds and expects to fail; ok_control builds normally so a broken harness fails loudly rather than passing vacuously.
• CI note: this PR's checks will only run once R1-02 (the CI workflow) is on main — currently blocked on the workflow OAuth scope (see the R1-02 status). Locally verified on macOS/AppleClang 17 under debug + asan-ubsan + -Werror.