lmxopcua/src/Drivers/ZB.MOM.WW.OtOpcUa.Driver.Sql.Browser/SqlDriverBrowser.cs

using System.Data.Common;
using System.Text.Json;
using System.Text.Json.Serialization;
using Microsoft.Extensions.Logging;
using Microsoft.Extensions.Logging.Abstractions;
using ZB.MOM.WW.OtOpcUa.Commons.Browsing;
using ZB.MOM.WW.OtOpcUa.Driver.Sql.Contracts;

namespace ZB.MOM.WW.OtOpcUa.Driver.Sql.Browser;

/// <summary>
///     Opens a <b>transient</b> database connection from form-supplied JSON for the AdminUI schema picker,
///     and hands it to a <see cref="SqlBrowseSession"/>. The AdminUI's <c>BrowseSessionRegistry</c> (its idle
///     TTL reaper included) owns the returned session unchanged — nothing about the SQL browse needs a
///     registry of its own.
///     <para><b>Two ways to name the database, and only one of them is persisted.</b>
///     <c>connectionStringRef</c> is the deployed form: a name resolved <em>in the AdminUI process</em> from
///     the environment as <c>Sql__ConnectionStrings__&lt;ref&gt;</c> (design §4.4), so a deployed artifact
///     never carries credentials. <c>connectionString</c> is an ad-hoc, <b>session-only</b> literal an
///     operator may paste to browse a database that has no ref yet; it is used for this one connection and
///     then forgotten — there is deliberately no cached-config field on this type, and nothing writes it
///     anywhere.</para>
///     <para><b>Precedence: a pasted literal wins over a ref</b>, and the ref is then not even read. The
///     literal cannot arrive from a persisted blob — <see cref="SqlDriverConfigDto"/> has no such property, so
///     the driver factory would drop it — which makes its presence proof that an operator typed it just now.
///     A config carrying both logs a warning (naming the shadowed <em>ref</em>, never any connection text) so
///     the override is visible rather than silent.</para>
///     <para><b>Credential hygiene is the point of this type.</b> A connection string is a secret. It is
///     passed to the provider and to nothing else: it never reaches a log line, an exception message, a field,
///     or anything persisted. Because ADO.NET providers are free to echo connection-string content in their
///     own errors, every failure on the open path goes through <see cref="Sanitize"/> before it leaves this
///     class.</para>
/// </summary>
public sealed class SqlDriverBrowser : IDriverBrowser
{
    /// <summary>
    ///     Prefix of the environment variable a <c>connectionStringRef</c> resolves through — the
    ///     double-underscore form .NET configuration uses for the <c>Sql:ConnectionStrings:&lt;ref&gt;</c>
    ///     key path.
    /// </summary>
    public const string ConnectionStringEnvironmentPrefix = "Sql__ConnectionStrings__";

    /// <summary>
    ///     Hard cap on the connect phase, layered on the caller's token. The AdminUI already bounds each
    ///     browse call at 20 s; this only stops a pathological provider-side connect from outliving the
    ///     picker entirely.
    /// </summary>
    private static readonly TimeSpan ConnectBudget = TimeSpan.FromSeconds(30);

    /// <summary>
    ///     Connection-string keys whose <em>values</em> are redacted out of any provider error text. The
    ///     whole connection string is redacted unconditionally; these are the parts that could survive as a
    ///     fragment. Server / database names are deliberately <b>not</b> redacted — "server X was not found"
    ///     is the diagnostic the operator needs, and a hostname is not a credential.
    /// </summary>
    private static readonly string[] CredentialKeys =
    [
        "password", "pwd", "user id", "uid", "user", "username", "accesstoken", "access token",
    ];

    /// <summary>
    ///     Kept deliberately identical in shape to the driver factory's options (design §5.1): enum-valued
    ///     knobs are authored as their <em>names</em>, so the browser parses a given <c>DriverConfig</c> blob
    ///     exactly as the runtime factory does. <see cref="JsonStringEnumConverter"/> also accepts ordinals,
    ///     so a numeric config still binds.
    /// </summary>
    private static readonly JsonSerializerOptions JsonOpts = new()
    {
        UnmappedMemberHandling = JsonUnmappedMemberHandling.Skip,
        PropertyNameCaseInsensitive = true,
        Converters = { new JsonStringEnumConverter() },
    };

    private readonly Func<SqlProvider, ISqlDialect?> _dialectSelector;
    private readonly Func<string, string?> _environmentReader;
    private readonly ILogger<SqlDriverBrowser> _logger;

    /// <summary>Creates a browser. Every dependency has a production default, so DI may construct it bare.</summary>
    /// <param name="dialectSelector">
    ///     Maps an authored <see cref="SqlProvider"/> onto the dialect that browses it, or <see langword="null"/>
    ///     for a provider this build does not carry. Defaults to <see cref="DefaultDialectSelector"/>
    ///     (SqlServer only, in v1). A constructor parameter rather than a test hatch: it is also how a P2
    ///     provider gets added without touching this class.
    /// </param>
    /// <param name="environmentReader">
    ///     Reads one environment variable. Defaults to <see cref="Environment.GetEnvironmentVariable(string)"/>.
    ///     Injectable because environment variables are process-global, and a test that sets one races every
    ///     other test in the assembly.
    /// </param>
    /// <param name="logger">Optional; defaults to <see cref="NullLogger{T}"/>.</param>
    public SqlDriverBrowser(
        Func<SqlProvider, ISqlDialect?>? dialectSelector = null,
        Func<string, string?>? environmentReader = null,
        ILogger<SqlDriverBrowser>? logger = null)
    {
        _dialectSelector = dialectSelector ?? DefaultDialectSelector;
        _environmentReader = environmentReader ?? Environment.GetEnvironmentVariable;
        _logger = logger ?? NullLogger<SqlDriverBrowser>.Instance;
    }

    /// <inheritdoc />
    /// <remarks>
    ///     Sourced from <see cref="SqlDriver.DriverTypeName"/>, the driver's interim local constant, rather
    ///     than from <c>DriverTypeNames</c> — the shared constant set is added by the driver-factory task,
    ///     because adding a member there reddens <c>DriverTypeNamesGuardTests</c> until a factory is
    ///     registered against it. Repointing that one constant repoints this too.
    /// </remarks>
    public string DriverType => SqlDriver.DriverTypeName;

    /// <summary>The environment variable a <c>connectionStringRef</c> resolves through.</summary>
    /// <param name="connectionStringRef">The authored reference name.</param>
    /// <returns>The full environment-variable name, e.g. <c>Sql__ConnectionStrings__MesStaging</c>.</returns>
    public static string EnvironmentVariableFor(string connectionStringRef) =>
        ConnectionStringEnvironmentPrefix + connectionStringRef;

    /// <summary>
    ///     The providers this build can browse. v1 constructs SQL Server only; every other
    ///     <see cref="SqlProvider"/> member is a reserved name, so it resolves to <see langword="null"/> and
    ///     the caller reports it as unavailable rather than crashing on a null dialect.
    /// </summary>
    /// <param name="provider">The authored provider.</param>
    /// <returns>The dialect, or <see langword="null"/> when this build does not carry the provider.</returns>
    public static ISqlDialect? DefaultDialectSelector(SqlProvider provider) =>
        provider == SqlProvider.SqlServer ? new SqlServerDialect() : null;

    /// <inheritdoc />
    /// <exception cref="InvalidOperationException">
    ///     The configuration is absent / malformed, names a provider this build does not carry, names neither
    ///     a <c>connectionStringRef</c> nor a literal, resolves a ref that is not set in the environment, or
    ///     the connection could not be opened.
    /// </exception>
    public async Task<IBrowseSession> OpenAsync(string configJson, CancellationToken cancellationToken)
    {
        if (string.IsNullOrWhiteSpace(configJson))
        {
            throw new InvalidOperationException(
                "The Sql browser requires a driver configuration; none was supplied.");
        }

        var literal = ReadPastedConnectionString(configJson);
        var config = Deserialize(configJson, literal);

        // Provider first: "this build cannot browse Postgres" is true regardless of how the database is
        // named, and answering it before touching the environment keeps the two failures from masking.
        var dialect = _dialectSelector(config.Provider)
            ?? throw new InvalidOperationException(
                $"Sql provider '{config.Provider}' is not available in this build; v1 browses "
                + $"'{SqlProvider.SqlServer}' only.");

        var reference = config.ConnectionStringRef;
        var connectionString = ResolveConnectionString(literal, reference);

        _logger.LogInformation(
            "AdminUI Sql browse session opening (provider {Provider}, connection from {Source})",
            config.Provider,
            DescribeSource(literal, reference));

        return await OpenSessionAsync(dialect, connectionString, cancellationToken).ConfigureAwait(false);
    }

    /// <summary>
    ///     Opens one transient connection and hands ownership to the session. <b>The session closes the
    ///     connection on its own disposal</b>, so the only disposal this method owns is the failure path —
    ///     disposing on success would double-close and leave the picker with a dead session.
    /// </summary>
    private static async Task<IBrowseSession> OpenSessionAsync(
        ISqlDialect dialect, string connectionString, CancellationToken cancellationToken)
    {
        var connection = dialect.Factory.CreateConnection()
            ?? throw new InvalidOperationException(
                $"The ADO.NET provider for '{dialect.Provider}' returned no connection object.");

        try
        {
            connection.ConnectionString = connectionString;

            using var connectCts = CancellationTokenSource.CreateLinkedTokenSource(cancellationToken);
            connectCts.CancelAfter(ConnectBudget);
            await connection.OpenAsync(connectCts.Token).ConfigureAwait(false);

            return new SqlBrowseSession(connection, dialect);
        }
        catch (Exception ex)
        {
            try { await connection.DisposeAsync().ConfigureAwait(false); }
            catch { /* best-effort: the original failure is the useful one. */ }

            if (ex is OperationCanceledException) throw;
            throw Sanitize(ex, connectionString);
        }
    }

    /// <summary>
    ///     Picks the connection string, literal first. A blank literal is treated as absent (an untouched
    ///     form field), never as an empty connection string.
    /// </summary>
    private string ResolveConnectionString(string? literal, string? reference)
    {
        if (!string.IsNullOrWhiteSpace(literal))
        {
            if (!string.IsNullOrWhiteSpace(reference))
            {
                _logger.LogWarning(
                    "Sql browse: a pasted connection string was supplied alongside connectionStringRef "
                    + "'{Ref}'. The pasted value wins for this session only and is not persisted.",
                    reference);
            }

            return literal;
        }

        if (string.IsNullOrWhiteSpace(reference))
        {
            throw new InvalidOperationException(
                "The Sql browser needs a database to browse: set 'connectionStringRef' to a name resolved "
                + $"from the environment as '{ConnectionStringEnvironmentPrefix}<ref>', or paste a "
                + "connection string into 'connectionString' for an ad-hoc browse.");
        }

        var variable = EnvironmentVariableFor(reference);
        var resolved = _environmentReader(variable);
        if (string.IsNullOrWhiteSpace(resolved))
        {
            // Name the exact variable: "the ref did not resolve" is unactionable, and the operator's next
            // move is to set this one name on the AdminUI host.
            throw new InvalidOperationException(
                $"connectionStringRef '{reference}' does not resolve: environment variable "
                + $"'{variable}' is not set on the AdminUI host. Set it there (the AdminUI resolves browse "
                + "connection strings in its own process), or paste a connection string for an ad-hoc "
                + "browse.");
        }

        return resolved;
    }

    /// <summary>How the connection string was obtained, for the log. Carries no connection text.</summary>
    private static string DescribeSource(string? literal, string? reference) =>
        !string.IsNullOrWhiteSpace(literal)
            ? "a pasted literal (session-only, not persisted)"
            : $"connectionStringRef '{reference}'";

    /// <summary>
    ///     Pulls the ad-hoc <c>connectionString</c> literal out of the raw JSON. It is read here rather than
    ///     off the DTO because <see cref="SqlDriverConfigDto"/> deliberately has no such property — the
    ///     persisted contract must not be able to carry a credential — and the DTO's
    ///     <see cref="JsonUnmappedMemberHandling.Skip"/> would silently drop it.
    /// </summary>
    /// <returns>The literal, or <see langword="null"/> when absent, blank, or not a JSON string.</returns>
    private static string? ReadPastedConnectionString(string configJson)
    {
        using var document = ParseDocument(configJson);
        if (document.RootElement.ValueKind != JsonValueKind.Object)
        {
            throw new InvalidOperationException(
                "The Sql browser's driver configuration must be a JSON object.");
        }

        foreach (var property in document.RootElement.EnumerateObject())
        {
            if (!string.Equals(property.Name, "connectionString", StringComparison.OrdinalIgnoreCase))
                continue;

            if (property.Value.ValueKind != JsonValueKind.String) return null;
            var value = property.Value.GetString();
            return string.IsNullOrWhiteSpace(value) ? null : value;
        }

        return null;
    }

    /// <summary>
    ///     Parses the raw configuration. The <see cref="JsonException"/> is <b>not</b> attached or echoed:
    ///     at this point nothing has been extracted, so there is no known secret to redact against, and the
    ///     malformed text may itself be a pasted connection string.
    /// </summary>
    private static JsonDocument ParseDocument(string configJson)
    {
        try
        {
            return JsonDocument.Parse(configJson);
        }
        catch (JsonException)
        {
            throw new InvalidOperationException(
                "The Sql browser's driver configuration is not valid JSON. (The parser's message is "
                + "withheld because the malformed text may carry a connection string.)");
        }
    }

    /// <summary>
    ///     Binds the typed configuration. A bind failure (e.g. an unknown <c>provider</c> name) is reported
    ///     with the parser's own message redacted against the literal, which is known by now.
    /// </summary>
    private static SqlDriverConfigDto Deserialize(string configJson, string? literal)
    {
        try
        {
            return JsonSerializer.Deserialize<SqlDriverConfigDto>(configJson, JsonOpts)
                ?? throw new InvalidOperationException(
                    "The Sql browser's driver configuration deserialized to null.");
        }
        catch (JsonException ex)
        {
            throw new InvalidOperationException(
                "The Sql browser's driver configuration could not be bound: "
                + Redact(ex.Message, CollectSecrets(literal)));
        }
    }

    /// <summary>
    ///     Turns a provider-side open failure into an exception that cannot carry the connection string.
    ///     <para><b>Live-probed, not assumed.</b> <c>Microsoft.Data.SqlClient</c> and
    ///     <c>Microsoft.Data.Sqlite</c> keep connection-string <em>values</em> out of their
    ///     <c>SqlException</c>/<c>SqliteException</c> text (a failed connect says "the server was not found",
    ///     never what it was handed). Their connection-string <b>parser</b> is the exception: it reports an
    ///     unrecognised keyword by quoting it, and a value carrying an unquoted <c>;</c> — legal inside a
    ///     password, which ADO.NET expects you to quote — splits, so the tail of the password is parsed as a
    ///     keyword and echoed verbatim. <c>Password=Sup3r;SecretTail</c> yields
    ///     <c>Keyword not supported: 'secrettail;connect timeout'.</c> — the credential, in the message,
    ///     lower-cased and therefore invisible to a naive ordinal search for the value.</para>
    ///     <para>So the parser's message (an <see cref="ArgumentException"/>) is <b>never</b> surfaced. Every
    ///     other failure is additionally passed through the substring redactor as a backstop, and when that
    ///     fires anywhere in the exception <em>chain</em> the inner exception is dropped too — keeping it
    ///     would re-expose through <c>ToString()</c> exactly what the message just removed.</para>
    /// </summary>
    private static InvalidOperationException Sanitize(Exception ex, string connectionString)
    {
        if (ex is ArgumentException)
        {
            return new InvalidOperationException(
                "The Sql browser could not open a connection: the connection string is malformed, or "
                + "carries a keyword this provider does not support. (The provider's own message is "
                + "withheld because it quotes connection-string text verbatim; check for an unquoted ';' "
                + "or '=' inside a value such as the password.)");
        }

        var secrets = CollectSecrets(connectionString);
        var leaked = ContainsAny(ex.ToString(), secrets);
        var reason = Redact(ex.Message, secrets);

        var message = leaked
            ? "The Sql browser could not open a connection: " + reason
                + " (the provider's error echoed the connection string, so it was redacted and the inner "
                + "exception withheld)"
            : "The Sql browser could not open a connection: " + reason;

        return new InvalidOperationException(message, leaked ? null : ex);
    }

    /// <summary>
    ///     The substrings that must never leave this class: the whole connection string, and the values of
    ///     its credential-bearing keys (which could surface on their own).
    /// </summary>
    private static List<string> CollectSecrets(string? connectionString)
    {
        var secrets = new List<string>();
        if (string.IsNullOrWhiteSpace(connectionString)) return secrets;

        secrets.Add(connectionString);

        try
        {
            var builder = new DbConnectionStringBuilder { ConnectionString = connectionString };
            foreach (var key in CredentialKeys)
            {
                if (!builder.TryGetValue(key, out var value)) continue;
                var text = value?.ToString();
                if (!string.IsNullOrWhiteSpace(text)) secrets.Add(text);
            }
        }
        catch (ArgumentException)
        {
            // A malformed connection string has no parseable parts; the whole-string entry still stands.
        }

        // Longest first, so redacting a short credential value cannot chop a longer secret into
        // unredactable halves.
        secrets.Sort(static (left, right) => right.Length.CompareTo(left.Length));
        return secrets;
    }

    private static bool ContainsAny(string text, List<string> secrets)
    {
        foreach (var secret in secrets)
        {
            if (text.Contains(secret, StringComparison.OrdinalIgnoreCase)) return true;
        }

        return false;
    }

    private static string Redact(string text, List<string> secrets)
    {
        foreach (var secret in secrets)
        {
            text = text.Replace(secret, "[redacted]", StringComparison.OrdinalIgnoreCase);
        }

        return text;
    }
}