lmxopcua/src/ZB.MOM.WW.OtOpcUa.Driver.AbLegacy/Import/RsLogixSymbolImport.cs

using System.Globalization;
using System.IO;
using System.Text;
using Microsoft.Extensions.Logging;
using Microsoft.Extensions.Logging.Abstractions;

namespace ZB.MOM.WW.OtOpcUa.Driver.AbLegacy.Import;

/// <summary>
///     Materialises <see cref="AbLegacyTagDefinition"/> entries from RSLogix 500 / 5
///     "Database Export" CSV. The expected column shape is
///     <c>Symbol,Address,Description,DataType,Scope</c> — a slight superset of what RSLogix
///     itself emits ("DataType" is RSLogix-supplied for symbol exports but ignored here in
///     favour of the file-letter prefix on <c>Address</c>; it is left in the schema for
///     forward-compatibility with editor tools that prefer to drive the type explicitly).
/// </summary>
/// <remarks>
///     <para>
///         The parser is deliberately tolerant: header row + comment lines (starting with
///         <c>;</c> or <c>#</c>) are skipped silently, headers are matched case-insensitively,
///         and quoted fields handle embedded commas the way RFC 4180 prescribes ("foo,bar"
///         → <c>foo,bar</c>; doubled quotes inside a quoted field collapse to a single
///         literal quote).
///     </para>
///     <para>
///         Type resolution defers to <see cref="AbLegacyAddress.TryParse(string?)"/> +
///         <see cref="TryResolveDataType"/> so the whole "what kind of file is N7?" knowledge
///         lives in one place. Function-file (<c>RTC</c>, <c>HSC</c>, …) and structure-file
///         (<c>PD</c>, <c>MG</c>, <c>PLS</c>, <c>BT</c>) prefixes are accepted but parsed
///         conditionally on <see cref="PlcFamilies.AbLegacyPlcFamily"/>; for the import path
///         we don't yet know the family so we use Slc500 as the parser context — that family
///         covers every common letter <see cref="RsLogixSymbolImport"/> needs to classify.
///     </para>
///     <para>
///         <see cref="System.IO.InvalidDataException"/> surfaces only when
///         <see cref="ImportOptions.IgnoreInvalid"/> is <c>false</c> — the default permissive
///         path logs a warning per malformed row and bumps the <c>SkippedCount</c> /
///         <c>ErrorCount</c> totals on <see cref="RsLogixImportResult"/>.
///     </para>
/// </remarks>
public sealed class RsLogixSymbolImport : IRsLogixImporter
{
    private readonly ILogger<RsLogixSymbolImport> _logger;

    public RsLogixSymbolImport() : this(NullLogger<RsLogixSymbolImport>.Instance) { }

    public RsLogixSymbolImport(ILogger<RsLogixSymbolImport> logger)
    {
        _logger = logger ?? NullLogger<RsLogixSymbolImport>.Instance;
    }

    /// <inheritdoc />
    public RsLogixImportResult Parse(Stream stream, string deviceHostAddress, ImportOptions? options = null)
    {
        ArgumentNullException.ThrowIfNull(stream);
        ArgumentException.ThrowIfNullOrWhiteSpace(deviceHostAddress);
        var opts = options ?? new ImportOptions();

        var tags = new List<AbLegacyTagDefinition>();
        var parsed = 0;
        var skipped = 0;
        var errors = 0;

        // detectEncodingFromByteOrderMarks=true honours UTF-8 BOMs (RSLogix tools on Windows
        // emit them often) without making the caller reach for a pre-decoded TextReader.
        // leaveOpen=true lets the caller manage the stream's lifecycle.
        using var reader = new StreamReader(stream, Encoding.UTF8, detectEncodingFromByteOrderMarks: true, bufferSize: 4096, leaveOpen: true);

        int? symbolIdx = null;
        int? addressIdx = null;
        int? descriptionIdx = null;
        int? dataTypeIdx = null;
        int? scopeIdx = null;
        var headerSeen = false;
        var lineNumber = 0;

        string? line;
        while ((line = reader.ReadLine()) is not null)
        {
            lineNumber++;
            if (string.IsNullOrWhiteSpace(line)) continue;
            var trimmed = line.TrimStart();
            if (trimmed.StartsWith(';') || trimmed.StartsWith('#')) continue;

            var fields = SplitCsv(line);
            if (fields.Count == 0) continue;

            if (!headerSeen)
            {
                // First non-blank, non-comment row — treat as header. Map every column we
                // recognise; missing required columns short-circuit the whole run with a
                // single InvalidDataException because the failure is structural, not
                // per-row.
                for (var i = 0; i < fields.Count; i++)
                {
                    var header = fields[i].Trim().ToLowerInvariant();
                    switch (header)
                    {
                        case "symbol": symbolIdx = i; break;
                        case "address": addressIdx = i; break;
                        case "description": descriptionIdx = i; break;
                        case "datatype":
                        case "data type":
                        case "type": dataTypeIdx = i; break;
                        case "scope": scopeIdx = i; break;
                    }
                }

                if (symbolIdx is null || addressIdx is null)
                {
                    throw new InvalidDataException(
                        $"RSLogix import header at line {lineNumber} is missing required Symbol or Address column. " +
                        $"Got: {string.Join(",", fields)}");
                }
                headerSeen = true;
                continue;
            }

            if (opts.MaxRowsToImport is int cap && parsed >= cap)
            {
                _logger.LogWarning(
                    "RSLogix import hit MaxRowsToImport={Cap} at line {LineNumber}; remaining rows skipped.",
                    cap, lineNumber);
                break;
            }

            // Per-row error scoping — we want a single bad row to skip cleanly without
            // dropping the rest of the file. The else branch in IgnoreInvalid=false mode
            // re-throws to surface the failure to the caller.
            try
            {
                // symbolIdx + addressIdx are guaranteed non-null past the header gate above.
                var symbol = SafeField(fields, symbolIdx!.Value);
                var address = SafeField(fields, addressIdx!.Value);
                var description = descriptionIdx.HasValue ? SafeField(fields, descriptionIdx.Value) : null;
                var scope = scopeIdx.HasValue ? SafeField(fields, scopeIdx.Value) : null;

                if (string.IsNullOrWhiteSpace(symbol) || string.IsNullOrWhiteSpace(address))
                {
                    skipped++;
                    _logger.LogWarning(
                        "RSLogix CSV row at line {LineNumber} skipped — missing Symbol or Address (symbol='{Symbol}', address='{Address}').",
                        lineNumber, symbol, address);
                    continue;
                }

                // Scope filter: row's Scope (or "Global" when blank) must match the filter
                // case-insensitively. RSLogix CSV scope values look like "Global" or
                // "Local:N" / "LOCAL:1" depending on the tool that emitted them.
                if (opts.ScopeFilter is { } wanted)
                {
                    var actual = string.IsNullOrWhiteSpace(scope) ? "Global" : scope.Trim();
                    if (!string.Equals(actual, wanted.Trim(), StringComparison.OrdinalIgnoreCase))
                    {
                        skipped++;
                        continue;
                    }
                }

                if (!TryResolveDataType(address.Trim(), out var dataType))
                {
                    if (!opts.IgnoreInvalid)
                    {
                        throw new InvalidDataException(
                            $"RSLogix CSV row at line {lineNumber} has unrecognised PCCC address '{address}'.");
                    }
                    errors++;
                    _logger.LogWarning(
                        "RSLogix CSV row at line {LineNumber} skipped — unrecognised PCCC address '{Address}'.",
                        lineNumber, address);
                    continue;
                }

                // Description column is parsed but currently unused — AbLegacyTagDefinition
                // doesn't carry a Description field today (the v2 schema ledger lives on the
                // server's metadata side of the bridge per #248). We retain the column in the
                // CSV header contract so a future schema bump can pick it up without breaking
                // existing exports. _ discard suppresses the unused-local warning.
                _ = description;
                tags.Add(new AbLegacyTagDefinition(
                    Name: symbol.Trim(),
                    DeviceHostAddress: deviceHostAddress,
                    Address: address.Trim(),
                    DataType: dataType,
                    Writable: true));
                parsed++;
            }
            catch (InvalidDataException) when (opts.IgnoreInvalid)
            {
                errors++;
                _logger.LogWarning("RSLogix CSV row at line {LineNumber} skipped — invalid data.", lineNumber);
            }
            catch (Exception ex) when (opts.IgnoreInvalid)
            {
                errors++;
                _logger.LogWarning(ex, "RSLogix CSV row at line {LineNumber} skipped — parser threw.", lineNumber);
            }
        }

        if (!headerSeen)
        {
            // Empty CSV (only blanks / comments) — return an empty result rather than
            // surface a "no header found" error. The CLI will report parsed=0 which is the
            // honest answer.
            return new RsLogixImportResult([], 0, skipped, errors);
        }

        return new RsLogixImportResult(tags, parsed, skipped, errors);
    }

    /// <summary>
    ///     Resolve a PCCC <paramref name="address"/> to the matching
    ///     <see cref="AbLegacyDataType"/>. Returns <c>false</c> for unparsable addresses.
    /// </summary>
    /// <remarks>
    ///     The mapping follows the file-letter table on
    ///     <see cref="AbLegacyAddress"/> doc comments:
    ///     N→Int, F→Float, B→Bit, L→Long, ST→String, T→TimerElement, C→CounterElement,
    ///     R→ControlElement, A→AnalogInt, S/I/O→Int (status / I/O bits resolve as Bit when
    ///     the address carries a <c>/N</c> bit suffix), PD→PidElement, MG→MessageElement,
    ///     PLS→PlsElement, BT→BlockTransferElement, function-file letters (RTC/HSC/etc.) →
    ///     MicroLogixFunctionFile.
    /// </remarks>
    public static bool TryResolveDataType(string address, out AbLegacyDataType dataType)
    {
        dataType = AbLegacyDataType.Int;
        // Use Slc500 as the parser family — it accepts every common letter the importer
        // sees in the wild. Family-specific gating (PLC-5 octal I:/O:, PD/MG/PLS/BT) only
        // matters for runtime addressing, not for shape classification at import time.
        var parsed = AbLegacyAddress.TryParse(address, PlcFamilies.AbLegacyPlcFamily.Plc5)
                  ?? AbLegacyAddress.TryParse(address, PlcFamilies.AbLegacyPlcFamily.Slc500)
                  ?? AbLegacyAddress.TryParse(address, PlcFamilies.AbLegacyPlcFamily.MicroLogix);
        if (parsed is null) return false;

        var letter = parsed.FileLetter;
        // Bit-within-word references on N/L/I/O/S files surface as Bit regardless of the
        // base file type. B-file references with no bit suffix are rare in real exports
        // but still classify as Bit (the wire-level element is a single word — Rockwell
        // convention is one bool per word).
        if (parsed.BitIndex is not null)
        {
            dataType = AbLegacyDataType.Bit;
            return true;
        }

        dataType = letter switch
        {
            "N" => AbLegacyDataType.Int,
            "F" => AbLegacyDataType.Float,
            "B" => AbLegacyDataType.Bit,
            "L" => AbLegacyDataType.Long,
            "ST" => AbLegacyDataType.String,
            "T" => AbLegacyDataType.TimerElement,
            "C" => AbLegacyDataType.CounterElement,
            "R" => AbLegacyDataType.ControlElement,
            "A" => AbLegacyDataType.AnalogInt,
            "I" or "O" or "S" => AbLegacyDataType.Int,
            "PD" => AbLegacyDataType.PidElement,
            "MG" => AbLegacyDataType.MessageElement,
            "PLS" => AbLegacyDataType.PlsElement,
            "BT" => AbLegacyDataType.BlockTransferElement,
            _ when AbLegacyAddress.IsFunctionFileLetter(letter) => AbLegacyDataType.MicroLogixFunctionFile,
            _ => AbLegacyDataType.Int,
        };
        return true;
    }

    private static string SafeField(IReadOnlyList<string> fields, int idx) =>
        idx >= 0 && idx < fields.Count ? fields[idx] : string.Empty;

    /// <summary>
    ///     RFC 4180-ish CSV splitter — quoted fields, doubled-quote escape, embedded comma
    ///     inside quoted fields. Avoids a third-party CSV dependency for a five-column
    ///     parser.
    /// </summary>
    internal static List<string> SplitCsv(string line)
    {
        var fields = new List<string>();
        var sb = new StringBuilder(line.Length);
        var inQuotes = false;
        for (var i = 0; i < line.Length; i++)
        {
            var c = line[i];
            if (inQuotes)
            {
                if (c == '"')
                {
                    // Doubled quote inside a quoted field is a literal `"`; otherwise the
                    // quote terminates the quoted segment.
                    if (i + 1 < line.Length && line[i + 1] == '"')
                    {
                        sb.Append('"');
                        i++;
                    }
                    else
                    {
                        inQuotes = false;
                    }
                }
                else
                {
                    sb.Append(c);
                }
            }
            else
            {
                switch (c)
                {
                    case '"':
                        inQuotes = true;
                        break;
                    case ',':
                        fields.Add(sb.ToString());
                        sb.Clear();
                        break;
                    default:
                        sb.Append(c);
                        break;
                }
            }
        }
        fields.Add(sb.ToString());
        return fields;
    }
}