matrix-gsheets-0.2.0

Breaking Changes

• Renamed authentication classes: BqAuthenticator → GsAuthenticator, BqAuthUtils → GsAuthUtils
  • See docs/0.1-0.2-MIGRATION.md for detailed migration guide with before/after examples

New Features

• GsheetsWriter.update(String, String, Matrix, ...): Write Matrix data to an existing spreadsheet and range
• GsheetsWriter.spreadsheetUrl(String): Convenience helper that returns the Google Sheets edit URL for a spreadsheet ID

Improvements

• Eliminated duplicated service-setup code: Extracted buildSheetsService() helpers in both GsheetsReader and GsheetsWriter
• Eliminated duplicated utilities in GsheetsWriter: Removed sanitizeSheetName(), toCell(), and MAX_SHEET_NAME_LENGTH — now delegates to GsUtil
• Fixed sheet-name extraction: When a range has no ! prefix (e.g., A1:D10), the matrix name defaults to '' instead of the raw range string
• Narrowed exception handling: GsUtil.getSheetNames(String, Sheets) now catches IOException specifically and declares throws IOException
• Improved numeric precision: GsConverter.asLocalDate(Number) uses longValue() instead of intValue() for day counts
• Idiomatic Groovy cleanup: Removed ~20 unnecessary return keywords from final expressions; replaced Java-style new ArrayList<>() + .add() with [] literals and << / collect
• Code quality: Removed commented-out tokeninfo URL from GsAuthUtils; converted inline usage docs to proper GroovyDoc
• Fixed GsAuthenticator log indentation: log.info call is now correctly inside the if (verbose) block
• Fixed static field ordering: Moved PROP_USER_HOME above ADC_FILE_PATH to eliminate initialization-order hazard
• Added missing return type declarations: GsConverter.asLocalTime(Object) → LocalTime, GsConverter.asSerial(Date) → BigDecimal
• Added historical-naming GroovyDoc notes to GsAuthenticator and GsAuthUtils explaining the Bq prefix

Test Changes

• Added 5 positive-case tests for GsConverter.toSerials (LocalDates, LocalDateTimes, LocalTimes, mixed types, empty list)
• Added 5 validation tests for GsheetsWriter.update() (null spreadsheetId, null range, null matrix, empty matrix, no rows)
• Moved 16 tests (sanitizeSheetName and toCell coverage) from GsExporterTest to GsUtilTest where they belong
• Removed brittle reflection-based tests from GsheetsReaderTest and GsheetsWriterTest

Dependency Updates

• com.google.apis:google-api-services-drive v3-rev20260220-2.0.0 → v3-rev20260322-2.0.0
• com.google.apis:google-api-services-sheets v4-rev20251110-2.0.0 → v4-rev20260213-2.0.0
• org.mockito:mockito-core 5.22.0 → 5.23.0
• org.mockito:mockito-junit-jupiter 5.22.0 → 5.23.0
• se.alipsa.nexus-release-plugin:se.alipsa.nexus-release-plugin.gradle.plugin 2.1.1 → 2.1.2

Matrix Avro 0.3.0

• matrix-avro now compiles Groovy statically by default via config/groovy/compileStatic.groovy; explicit @CompileStatic annotations were removed.
• CodeNarc is now enforced for the module with ignoreFailures = false, and warnings were fixed.
• Schema inference was hardened/refactored through ColumnProfile, especially for mutable/object-typed columns, decimal metadata, lists, maps, and record-like maps.
• fixed pre-epoch local-timestamp-millis reads by using floor modulo for nanosecond remainders
• made explicit timestamp-millis writes of LocalDateTime timezone-stable by interpreting them at UTC
• aligned stream ownership with the public docs
  • InputStream read/schema overloads leave caller-owned streams open
  • OutputStream write overloads leave caller-owned streams open
• removed the writer schema cache so mutated matrices cannot reuse stale schemas
• added public schema inspection APIs through MatrixAvroReader.schema(...)
• added convenience factories and shortcuts
  • AvroReadOptions.defaults() and AvroReadOptions.named(...)
  • AvroWriteOptions.defaults() and AvroWriteOptions.exactDecimals()
  • MatrixAvroWriter.writeExactDecimals(...) and writeExactDecimalBytes(...)
  • AvroSchemaDecl.decimalColumn(...), arrayOf(...), and mapOf(...)
• tightened public validation for schema building and null options
• refreshed README, tutorial, and cookbook examples for schema inspection and decimal-safe writes

Matrix SQL v2.4.0

SQL identifier handling

• New SqlIdentifier utility class for safe quoting and rendering of table/column names containing spaces, mixed case, reserved words, punctuation, or embedded quotes.
• All generated SQL (DDL, INSERT, UPDATE, DROP) now routes through SqlIdentifier, replacing ad-hoc string concatenation with proper double-quote escaping.
• MatrixDbUtil.tableName() sanitises matrix names with a regex-based normaliser instead of targeted character replacements.

Prepared-statement convenience APIs

• select(String, List, String) — parameterised SELECT returning a Matrix.
• update(String, List) — parameterised UPDATE/INSERT/DELETE.
• delete(String, List) — parameterised DELETE.
• execute(String, List) — parameterised arbitrary SQL returning a Map<Integer, Object> of result sets and update counts.
• insert(String, Matrix) — insert a Matrix into an explicitly named table.
• update(String, Row) — overload that always throws IllegalArgumentException to prevent accidental unconstrained updates; use update(String, Row, String...) with match columns instead.

Managed (non-owning) connections

• New constructors MatrixSql(Connection, DataBaseProvider) and MatrixSql(Connection, SqlTypeMapper) wrap an externally supplied connection. close() leaves externally supplied connections open and usable.

Factory offline fallback

• MatrixSqlFactory.FALLBACK_VERSIONS map centralises pinned fallback versions for H2 and Derby.
• createH2(), createDerby(), and generic create() fall back to the pinned version when Maven Central is unreachable instead of throwing.
• Error message for unsupported providers now includes the dependency coordinates and a "no fallback version is configured" hint.

MatrixResultSet hardening

• Guards (ensureOpen, ensureCurrentRow, checkedColumnIndex) prevent operations on closed, unpositioned, or out-of-range result sets.
• All column accessors routed through readValue() helper for consistency.
• updateRow() is a documented no-op for detached result sets.
• Null-safe primitive getters return JDBC-specified defaults (0, false) when the value is null.
• unwrap() and isWrapperFor() follow the strict JDBC contract.
• Calendar-aware getDate, getTime, getTimestamp and getURL corrected.

MatrixDbUtil improvements

• Default minimum sizes for VARCHAR (255), DECIMAL precision (38) and scale (10) when column scanning finds no data.
• insert(Connection, String, Matrix, boolean) overload passes addQuotes through to generated SQL.

Build and dependency changes

• Remove dependency-resolver dependency; replaced by maven-utils which now covers the same functionality.
• Migrate inline dependency declarations to Gradle version catalog (libs.versions.toml).
• Add groovier-junit test dependency for Groovy-friendly JUnit assertions.
• Remove log4j-to-slf4j test runtime dependency.
• Dependency upgrades:
  • se.alipsa.groovy:data-utils [2.0.4 -> 2.0.6]
  • se.alipsa:maven-utils (replaces maven-3.9.11-utils 1.1.0) [-> 1.4.1]
  • commons-io:commons-io [2.21.0 -> 2.22.0]

Documentation and test hygiene

• Comprehensive GroovyDoc on all public methods and constructors.
• Expanded README with runnable examples for all public workflows.
• Replaced all println/System.err.println in tests with assertions or Logger.

Matrix Arff 0.2.1

• Fix nominal sentinel values (?, empty string, %-prefixed) being written unquoted, causing lossy ARFF round-trips
• Add ArffDateFormats utility to share strict (lenient=false), UTC, Locale.ROOT date formatter creation between reader and writer
• Fix date parsing to reject invalid dates (e.g. 2026-02-31) instead of silently normalizing them
• Fix published SCM URL in POM (matrix-arff/tree/master → tree/main/matrix-arff)
• Apply module-wide @CompileStatic via compileStatic.groovy build configuration
• Add regression tests for nominal sentinel round-trips, invalid date rejection, and UTC date output
• Fixed all codenarc warnings and change the build to fail on any new warnings.

Matrix Parquet v0.5.0

• Add SPI integration: MatrixParquetFormatProvider registers .parquet extension with Matrix SPI so Matrix.read(file) and matrix.write(file) work without explicit imports
• Add ParquetReadOptions and ParquetWriteOptions typed options classes with describe() and fromMap() for runtime discovery and SPI use
• Add builder API: MatrixParquetReader.builder() and MatrixParquetWriter.builder(matrix) as the recommended fluent interface
• Add write(OutputStream) and write(Path) overloads to writer; add read(byte[]), read(URL), read(Path), read(InputStream) to builder
• Add writeBytes() method to write to a byte array without a file
• Strengthen ParquetWriteOptions.validate(): enforce precision > 0, scale >= 0, scale ≤ precision for both uniform and per-column decimal meta
• Add decimalMeta per-column precision/scale validation in validateDecimalMeta (checks shape, range, and consistency)
• Fix bug: resource leak in MatrixParquetReader — reader now wrapped in withCloseable to ensure streams are always closed
• Fix bug: negative BigDecimal values padded with 0x00 instead of 0xFF causing sign bit corruption on read
• Fix bug: timestamp type mapped to MILLIS instead of MICROS causing precision loss
• Fix bug: deprecated BigDecimal.ROUND_HALF_UP replaced with RoundingMode.HALF_UP
• Fix hasUniformPrecisionAndScale() semantics: now returns true only when both precision and scale are non-null
• Enable @CompileStatic by default for all production sources
• Enable CodeNarc with ignoreFailures=false; fix all pre-existing violations
• Remove ivy dependency (was unused)
• Make internal APIs private; simplify readFromInputStream
• Migrate tests to @TempDir; remove manual temp-file cleanup boilerplate
• Correct POM publication URLs: url, license.url, and scm.url now use tree/main paths (Maven Central convention)
• Add README sections: "At a Glance" goal/entry-point table, "API Surface" reference, "Default Behavior" for naming and decimal defaults
• Upgrade dependencies
  • org.apache.hadoop:hadoop-common [3.4.2 -> 3.4.3]
  • org.apache.hadoop:hadoop-mapreduce-client-core [3.4.2 -> 3.4.3]

Matrix Json v2.2.0

• add fluent WriteBuilder API for JsonWriter: JsonWriter.write(matrix).indent().to(file)
• add matrix name derivation from file/URL in JsonReader
• add matrixName option to JsonReadOptions to override file-derived name
• add type-conversion support (types, dateTimeFormat) to JsonReadOptions / JsonFormatProvider
• add readString(String) convenience method to JsonReader
• add missing Path and String write overloads for the formatter variant of JsonWriter
• switch JsonWriter from Groovy's groovy.json.* to Jackson streaming API (O(columns) peak memory)
• remove groovy-json build dependency (Jackson used for both reading and writing)
• enable build-level @CompileStatic for all production code
• add module-level CodeNarc configuration with ignoreFailures = false
• fix NPE risk in writeString when columnFormatters is null
• fix resource leak in JsonReader.read(File, Charset) (nested withCloseable)
• fix missing directory/parent-dir guards in formatter write(File, ...) overload
• deprecate all existing static methods on JsonWriter in favor of fluent API
• change JSON float parsing to deserialize floating-point numbers as BigDecimal instead of Double so imported decimal values keep their exact textual precision
• upgrade dependencies
  • com.fasterxml.jackson.core:jackson-core 2.21.0 -> 2.21.2
  • com.fasterxml.jackson.core:jackson-databind 2.21.0 -> 2.21.2

Matrix Stats v2.4.0

Native runtime cleanup and idiomatic Groovy API expansion

Runtime and Dependency Changes

• Remove Apache Commons Math from the runtime dependency surface. commons-math3 is now test-only and used to validate native implementations.
• Add EJML Simple as an internal implementation dependency for the public linear algebra facade.
• Configure matrix-stats production Groovy compilation through the shared compile-static Gradle script.
• Make CodeNarc fail the build instead of reporting non-blocking findings.

New Public APIs

Linear Algebra

• Add se.alipsa.matrix.stats.linalg.Linalg for dense matrix operations:
  inverse, determinant, linear solves, real eigenvalues, and SVD.
• Add Matrix and Grid adapters with Groovy-facing return types:
  Matrix, Grid<BigDecimal>, BigDecimal, List<BigDecimal>, and SvdResult.
• Add LinalgSingularMatrixException and shared adapter utilities for singular and invalid matrix input handling.
• Add compatibility classes under se.alipsa.matrix.stats.linear for native matrix algebra support.

Linear Interpolation

• Add se.alipsa.matrix.stats.interpolation.Interpolation for public linear interpolation.
• Support explicit (x, y, targetX) interpolation, evenly spaced series interpolation, and Matrix/Grid-backed column interpolation.
• Reject extrapolation, unsorted domains, duplicate domain values, length mismatches, ragged grids, and non-numeric columns.
• Keep spline logic internal to formula/GAM smooth-term support; public spline interpolation is not part of the 2.4.0 API.

Formula and Model-Frame Pipeline

• Add R-style formula parsing and normalization support, including additive terms, intercept control, interactions, shorthand expansion, quoted identifiers, numeric transformations, poly(...), and s(...) smooth terms.
• Add ModelFrame and ModelFrameResult for design-matrix construction from Matrix data.
• Add a Groovy-native operator DSL for formulas using |, noIntercept, interaction(...), smooth(...), and I { ... }.
• Add categorical treatment encoding, dot expansion, subset support, NA handling, weights, offsets, and external environment variable resolution.
• Add NaAction and formula metadata classes to make downstream model fitting explicit.
• Reject unsupported response forms, smooth-term interactions, and unsupported frame metadata instead of silently ignoring them.

Fit Registry and Formula-Based Regression

• Add FitRegistry, FitMethod, FitOptions, and FitResult for named fit-method dispatch.
• Add built-in lm, loess, and gam fit methods.
• Add FitDsl convenience entry points so Groovy callers can write lm(data) { y | x + group }, loess(data) { y | x }, and gam(data) { y | smooth(time, 6) + group }.
• Add MultipleLinearRegression, LmMethod, LoessMethod, and GamMethod.
• Add LoessOptions and GamOptions with Groovy-facing numeric option surfaces.

Native Distributions

• Add native NormalDistribution, ChiSquaredDistribution, and HypergeometricDistribution.
• Extend TDistribution, FDistribution, and SpecialFunctions to remove Commons Math runtime usage.
• Add Groovy-facing Number overloads returning BigDecimal for public scalar distribution APIs.
• Add typed entry points for F-distribution ANOVA helpers to avoid unsafe generic overload dispatch.

Native Solvers and Optimization

• Replace Commons Math optimizer/solver runtime usage with native implementations.
• Add BrentSolver for one-dimensional bracketing root finding.
• Update GoalSeek to use the native Brent-Dekker solver.
• Add NelderMeadOptimizer for derivative-free multivariate minimization.
• Add LinearProgramSolver for equality-form linear programs with non-negative variables.
• Add UnivariateObjective and MultivariateObjective interfaces with Groovy-facing numeric bridges.

Idiomatic Groovy Numeric API Cleanup

• Move public scalar result values toward BigDecimal and public numeric inputs toward Number.
• Add NumericConversion to centralize finite-value checks, BigDecimal conversion, array/list conversion, alpha validation, and exact integer validation.
• Add StatUtils and LeastSquaresKernel for small internal double-precision kernels where performance or algorithm constraints justify primitive arrays.
• Replace duplicate numeric coercion and conversion logic across regression, interpolation, distributions, time-series, KDE, and solver code.
• Add typed compatibility entry points for Johansen and F-distribution list/array APIs where JVM erasure makes same-name generic overloads unsafe.
• Deprecate primitive/double-style identity accessors in selected result classes where Groovy-facing properties are now preferred.

Time-Series and Statistical Robustness

• Extract shared time-series utility logic into TimeSeriesUtils.
• Improve singular-matrix handling and error messages in time-series code.
• Harden Johansen critical-value handling for unsupported variable counts.
• Improve CCM library-size validation by rejecting non-integral numeric inputs instead of truncating silently.
• Update ANOVA and contingency result APIs to accept Number alpha inputs.

Tests and Quality

• Add coverage for formula parsing, model-frame construction, design matrices, spline basis expansion, fit registry, lm, loess, gam, multiple linear regression, interpolation, linalg, SVD, native distributions, native solvers, numeric conversion, and least-squares kernels.
• Add direct unit tests for GroupEstimator.estimateNumberOfGroups and estimateKByElbow, covering both the double[][] and Groovy-facing List overloads, custom maxK/iterations, and error cases (too few points, too few distinct points).
• Add benchmark-oriented tests for selected Groovy-facing paths versus retained primitive kernels.
• Update tests for BigDecimal/Groovy-friendly assertions and numeric API behavior.
• Increase coverage around null handling, weights, offsets, subset filtering, categorical encoding, exact integer validation, singular matrices, and solver convergence.

Documentation

• Refresh matrix-stats README, tutorial, and cookbook docs for the 2.4.0 API surface.
• Document the Commons Math runtime removal and EJML implementation detail.
• Add examples for Linalg, Interpolation, formula/model-frame fitting, the Groovy formula DSL, fit convenience helpers, native distributions, and native solvers.
• Refresh broader tutorial setup snippets to use the current BOM and Groovy versions.

Matrix Core v3.7.1

• Preserve row count for zero-column matrices created from empty-row input so
  downstream consumers can retain observation alignment even when a design
  matrix contains no predictor columns.
• Added missing PackageScope import to Matrix.groovy
• CodeNarc cleanup across ~25 main source files (mostly style, but includes making the metaData field final)

Matrix BigQuery v0.6.1

• Bug fix: Added isExpectedInsertAllPreconditionFailure() helper and downgraded expected InsertAll precondition failures from log.error → log.debug to reduce noise.
• Dependency updates:
  • com.google.auth:google-auth-library-bom 1.43.0 -> 1.46.0
  • com.google.auth:google-auth-library-oauth2-http 1.43.0 -> 1.46.0
  • com.google.cloud:google-cloud-bigquery 2.62.0 -> 2.65.0
  • com.google.cloud:google-cloud-bigquerystorage 3.24.0 -> 3.27.0
  • com.google.cloud:google-cloud-resourcemanager 1.90.0 -> 1.93.0
• Codenarc fixes, build now fails if any new violations are introduced

Matrix CSV v2.3.0

• clean up deprecated CsvImporter dead code and suppress legacy enum naming violations without breaking backward compatibility
• fix remaining CodeNarc priority 2 issues across matrix-csv, including builder field naming and deprecated Reader charset documentation
• make the typed options API first-class through CsvReadOptions / CsvWriteOptions direct overloads, SPI normalization via fromMap(...), and parity between direct and SPI usage
• correct the CSV format model so Excel and RFC4180 are no longer treated as the same preset
• make the write options contract explicit by supporting only observable write-time options and rejecting parse-only write settings at the typed/SPI boundary
• align Excel behavior with Apache Commons CSV semantics, including QuoteMode.ALL_NON_NULL, CRLF output, and missing-header support
• refresh the README, cookbook, and tutorial around CsvReader, CsvWriter, typed options, default behavior, and runtime option discovery