Add histogram axes, accumulators, and frequency views - #108
Merged
Merged
Conversation
AI-Assisted-By: Codex (GPT-6 Astra Medium)
AI-Assisted-By: Codex (GPT-6 Astra Medium)
AI-Assisted-By: Codex (GPT-6 Astra Medium)
AI-Assisted-By: Codex (GPT-6 Astra Medium)
AI-Assisted-By: Codex (GPT-6 Astra Medium)
AI-Assisted-By: Codex (GPT-6 Astra Medium)
AI-Assisted-By: Codex (GPT-6 Astra Medium)
AI-Assisted-By: Codex (GPT-6 Astra Medium)
AI-Assisted-By: Codex (GPT-6 Astra Medium)
Store read-only handles while preserving shared count updates and reference-counted ownership. Allow const views to be saved and sliced into mutable cursors, and add VariableAxis.lightConst support. Add documentation examples and regression tests for const access, storage lifetime, and range operations. All 49 test modules pass. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Add frequencyBins with live counts and frequencies, const access, slicing, and independent cursors. Isolate escape-check detection so borrowing requires @System when the necessary checks are unavailable. Co-locate bin views with their accumulators and extract shared internal helpers. Add documentation examples and behavior and lifetime tests. Enable the histogram suite in the DIP1000 build. Fix VariableAxis lookup safety by borrowing the search slice while retaining ownership in the axis. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Add cumulativeFrequency with a configurable floating-point output type, defaulting to double. Sum current counts through the selected bin, including underflow, and divide by the total including overflow. Return NaN when the total is zero. Make both histogram flow accessors const. Add documentation examples and tests covering output types, flow options, storage types, const access, insertion, merging, and invalid indices. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Add cumulativeFrequencies with a configurable floating-point output type, defaulting to double. Compute all ordinary-bin cumulative frequencies in one pass and return an independently owned, reference-counted Mir slice. Include underflow in cumulative counts and overflow in the total. Return NaN values when the total is zero. Add documentation examples and tests for flow handling, output types, const access, and snapshot independence after source updates and destruction. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Add a cumulativeFrequencies overload for writable floating-point arrays and one-dimensional Mir slices. Infer output precision from destination elements and require one element per ordinary bin. Reuse the destination overload when allocating snapshots. Document the requirement that output storage not overlap accumulator counts. Add documentation examples and tests for buffer reuse, strided slices, const sources, stack output, and invalid destinations. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Use shared boundary calculations for regular and transformed axes, preserving the supplied endpoints. Check estimated bin indices against the original observation and fall back to binary search when needed. Reject invalid grids and coincident adjacent boundaries during assertion-enabled construction. Add documentation examples and regression tests covering shared edges, extreme float/double/real values, interval closure, circular endpoints, and transformed-bin membership. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Reject nonpositive IntegralAxis bin counts and upper bounds that do not exceed the lower bound. For integral count and bin types, check that the count fits BinType and the endpoint addition cannot overflow. Require VariableAxis boundaries to be strictly increasing with a bin count that fits CountType. Preserve constant-time const conversion by reusing validated boundaries. Check that accumulator storage matches its axes. Document multiple-axis marginal histograms and add regression tests for invalid axes, integral endpoint limits, and mismatched storage. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Treat each argument as a separate observation for single-axis histograms and frequency accumulators. Require exactly one compatible coordinate per axis for multiple-axis histograms. Share observation-type checks and accept const and immutable numeric inputs. Update frequency totals after each successful insertion. Add compiler checks and regression tests for argument counts, invalid types, qualified inputs, category strings, flows, and partial batches. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Make underflow, overflow, and their relative-frequency accessors callable on const frequency accumulators. Add @nogc tests for histogram construction, insertion, merging, category handling, bin views, frequency reads, cumulative snapshots, caller-supplied output, and break selection. Check borrowed frequency-view traversal without GC allocation in both escape-checking modes. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Replace histogram-specific test guards with mir_stat_test and restore the full suite in default, coverage, and DIP1000 builds. Rename the DIP1000 lifetime-test flag to mir_stat_test_lifetime. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Move floating-to-count conversion into a helper that DMD does not inline, avoiding compiler crashes when callers compare bin indices with zero. Keep the rest of axis lookup inlineable. Add regression coverage for inlined lookup with float, double, and real inputs under both interval conventions. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Compare the real and imaginary components using a small tolerance instead of exact equality, accommodating rounding differences in optimized builds. AI-Assisted-By: Codex (GPT-6 Astra Medium)
AI-Assisted-By: Codex (GPT-6 Astra Medium)
Reject nonfinite bounds and collapsed floating-point bin boundaries. Use shared rounded boundaries for bin descriptions and lookup, while preserving direct integer lookup and the DMD conversion workaround. Add regression coverage for float, double, and real at precision limits and fractional origins, including both closure options and circular axes. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Use exact arithmetic for integer spans and integer widths, preserving distances between nearby large values. Handle overflowing floating-point spans and ratios that underflow to zero. Support integer arrays directly and validate inputs, widths, and count representability before conversion. Document constant-input behavior. Add regression coverage for extreme values, integer widths, invalid inputs, and CountType limits, including @nogc tests. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Correct parameter descriptions, sturges references, and documentation typos. Clarify that the remaining storage-shape work concerns joint histograms; marginal storage already has validation. Remove trailing whitespace from histogram modules while preserving CRLF line endings and final newlines. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Provide self-contained guidance for routine Mir development, numerical and performance work, and compiler investigations. Encourage reuse of established findings and proportional verification. Allow optional use of other available skills without requiring companion skills. Retain useful regression tests when removing compiler workarounds. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Check transform compatibility after BinType is deduced. Probe template transforms through the Mir invocation path so incompatible candidates are rejected without hard compilation errors. Add regression coverage for explicit and inferred types and inverses, axis options, boundary counts, and invalid transforms. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Move the elementCount import into the shared template scope so array inputs compile with both default and explicit count types. Add regression coverage for mutable and const arrays and agreement with slice inputs. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Wrap assignments in statement blocks inside __traits(compiles) to avoid parser errors while preserving checks that read-only data cannot be modified. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Import DefaultCountType within each documentation example that uses it so the examples compile independently of the axis module's imports. AI-Assisted-By: Codex (GPT-6 Astra Medium)
Add missing parameter descriptions and remove entries that do not match their overloads. Correct VariableAxis parameter names so histogram documentation builds with warnings treated as errors. AI-Assisted-By: Codex (GPT-6 Astra Medium)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Add histogram functionality to mir-stat, including axis construction,
incremental counting, and frequency calculations.
The implementation provides:
construction through
rchistogram.cumulative frequencies, and live frequency views.
Boundary validation and lookup account for floating-point rounding. Tests
cover extreme values, qualifiers, borrowed-view lifetimes, and
@nogcoperations. CI now includes DIP1000 and optimized release configurations.
Current limitations
Multiple axes record separate marginal counts, not joint multidimensional
bins. Joint storage and flow bins, weighted histograms, per-bin statistical
accumulators, and dynamically widening counters remain deferred.
Borrowed frequency views require their source accumulator to remain alive
and in place. Their safe use depends on compiler escape checking.
Validation
All six local configurations pass with 46 modules each:
Hosted Linux/Windows CI remains to be verified on the PR.
Assistance
Developed with AI assistance from Codex (GPT-6 Astra Medium).