fix: clear the analyzer warnings in src

[phmatray]

Clears every instance of the four warning classes CI reported. GitHub caps annotations at 10 per job, so the reported list was a subset — this covers all of them.

Warning	Reported	Actually present	Fixed
CS1587 — XML comment not on a valid element	3	9	9
CS1573 — missing <param> tag	4	4	4
IDE0290 — use primary constructor	1	1	1
IDE0046 — if can be simplified	2	2	2

CS1587 — nine result records

Each file ended with a documentation block describing the inherited Real property, sitting after the last member with no element to attach to.

The text was useful domain detail — the range DX takes, what a dominant-cycle period means — so it moved into the class-level <remarks> rather than being deleted:

/// <remarks>
/// The <see cref="SingleOutputResult.Real"/> array holds the Directional Movement Index values.
/// Values range from 0 to 100, where higher values indicate stronger trends (either up or down).
/// Values below 20 typically indicate weak trends, while values above 40 suggest strong trends.
/// </remarks>

CorrelResult's block only restated what its existing <remarks> already said, so that one is simply removed.

CS1573 — four float[] overloads

Each takes an optional parameter the documentation never mentioned. One of them was actively misleading: MacdFix claimed "Uses fixed values: fastPeriod=12, slowPeriod=26, signalPeriod=9" while signalPeriod was in fact a settable parameter. It now says the fast and slow periods are fixed and the signal period is adjustable.

IDE0290 — CandleIndicator<T>

Converted to a primary constructor, with the parameter documentation moved onto the type. The generated constructor signature is unchanged, so this is source and binary compatible — and it's the base class for all 61 candlestick patterns, so the 673 candle tests are the real check.

IDE0046 — ValidationHelper

ValidateIndexRange becomes a conditional cascade.

ValidateSingleInputIndicator was a hand-rolled guard chain. Rather than nest three ternaries to satisfy the analyzer, it now calls the ValidateAll helper this same class already exposes and the indicators already use (see Atr/TAFunc.cs):

return ValidateAll(
    () => ValidateIndexRange(startIdx, endIdx),
    () => ValidateArrays(inReal, outReal),
    () => optInTimePeriod.HasValue
        ? ValidatePeriodRange(optInTimePeriod.Value, minPeriod, maxPeriod)
        : Success);

One expression, still short-circuiting at the first failure, and lazier than the original — each check now only runs if the previous one passed.

Verification

dotnet build TaLibStandard.sln -c Release   ->  0 errors
grep for CS1587|CS1573|IDE0290|IDE0046 in src  ->  no matches

dotnet test TaLibStandard.sln -c Release
  Candles.UnitTests              673 passed
  Functions.UnitTests            232 passed
  Samples.Backtesting.UnitTests  220 passed
  ->  1125 passed, 0 failed

One thing to know

The generated pages under docs/functions/ and docs/common/ now lag these XML comment changes. I deliberately did not regenerate them: this machine emits different character encoding than the committed output (🡒 vs →), so regenerating would bury a 15-file change under ~175 spurious diffs — exactly what went wrong on #106.

They should be regenerated on a machine whose encoding matches, or the encoding should be pinned so this stops being a trap for contributors.

Worth noting on the IDE0046 rule specifically: .editorconfig line 83 sets dotnet_style_prefer_conditional_expression_over_return = true:silent, but AnalysisModeStyle=All in Directory.Build.props escalates it to a warning anyway. The .editorconfig's stated intent is being overridden. Worth reconciling those two if the style rules keep surfacing.

🤖 Generated with Claude Code