openrs2/archive/src/main/kotlin/org/openrs2/archive/cache/CacheImporter.kt

package org.openrs2.archive.cache

import io.netty.buffer.ByteBuf
import io.netty.buffer.ByteBufAllocator
import io.netty.buffer.ByteBufUtil
import org.openrs2.archive.container.Container
import org.openrs2.archive.container.ContainerImporter
import org.openrs2.buffer.use
import org.openrs2.cache.Js5Archive
import org.openrs2.cache.Js5Compression
import org.openrs2.cache.Js5Index
import org.openrs2.cache.Store
import org.openrs2.cache.VersionTrailer
import org.openrs2.crypto.Whirlpool
import org.openrs2.db.Database
import java.io.IOException
import java.sql.Connection
import java.sql.Types
import javax.inject.Inject
import javax.inject.Singleton

@Singleton
public class CacheImporter @Inject constructor(
    private val database: Database,
    private val alloc: ByteBufAllocator
) {
    private class Index(
        val archive: Int,
        val index: Js5Index,
        data: ByteBuf
    ) : Container(data) {
        override val encrypted: Boolean = false
    }

    private class Group(
        val archive: Int,
        val group: Int,
        data: ByteBuf,
        val version: Int,
        override val encrypted: Boolean
    ) : Container(data)

    public suspend fun import(store: Store) {
        database.execute { connection ->
            ContainerImporter.prepare(connection)

            // import indexes
            val indexes = mutableListOf<Index>()
            try {
                for (archive in store.list(Js5Archive.ARCHIVESET)) {
                    indexes += readIndex(store, archive)
                }

                val cacheId = addCache(connection, indexes)

                for (index in indexes) {
                    addIndex(connection, cacheId, index)
                }
            } finally {
                indexes.forEach(Index::release)
            }

            // import groups
            val groups = mutableListOf<Group>()
            try {
                for (archive in store.list()) {
                    if (archive == Js5Archive.ARCHIVESET) {
                        continue
                    }

                    for (id in store.list(archive)) {
                        val group = readGroup(store, archive, id) ?: continue
                        groups += group

                        if (groups.size >= BATCH_SIZE) {
                            addGroups(connection, groups)

                            groups.forEach(Group::release)
                            groups.clear()
                        }
                    }
                }

                if (groups.isNotEmpty()) {
                    addGroups(connection, groups)
                }
            } finally {
                groups.forEach(Group::release)
            }
        }
    }

    private fun readGroup(store: Store, archive: Int, group: Int): Group? {
        try {
            store.read(archive, group).use { buf ->
                val version = VersionTrailer.strip(buf) ?: return null
                val encrypted = Js5Compression.isEncrypted(buf.slice())
                return Group(archive, group, buf.retain(), version, encrypted)
            }
        } catch (ex: IOException) {
            return null
        }
    }

    private fun addGroups(connection: Connection, groups: List<Group>) {
        val containerIds = ContainerImporter.addContainers(connection, groups)

        connection.prepareStatement(
            """
            INSERT INTO groups (archive_id, group_id, container_id, truncated_version)
            VALUES (?, ?, ?, ?)
            ON CONFLICT DO NOTHING
        """.trimIndent()
        ).use { stmt ->
            for ((i, group) in groups.withIndex()) {
                stmt.setInt(1, group.archive)
                stmt.setInt(2, group.group)
                stmt.setLong(3, containerIds[i])
                stmt.setInt(4, group.version)
                stmt.addBatch()
            }

            stmt.executeBatch()
        }
    }

    private fun readIndex(store: Store, archive: Int): Index {
        return store.read(Js5Archive.ARCHIVESET, archive).use { buf ->
            Js5Compression.uncompress(buf.slice()).use { uncompressed ->
                Index(archive, Js5Index.read(uncompressed), buf.retain())
            }
        }
    }

    private fun addCache(connection: Connection, indexes: List<Index>): Long {
        val len = indexes.size * (1 + Whirlpool.DIGESTBYTES)
        val whirlpool = alloc.buffer(len, len).use { buf ->
            for (index in indexes) {
                buf.writeByte(index.archive)
                buf.writeBytes(index.whirlpool)
            }

            Whirlpool.whirlpool(ByteBufUtil.getBytes(buf, 0, buf.readableBytes(), false))
        }

        connection.prepareStatement(
            """
            SELECT id
            FROM caches
            WHERE whirlpool = ?
        """.trimIndent()
        ).use { stmt ->
            stmt.setBytes(1, whirlpool)

            stmt.executeQuery().use { rows ->
                if (rows.next()) {
                    return rows.getLong(1)
                }
            }
        }

        connection.prepareStatement(
            """
            INSERT INTO caches (whirlpool)
            VALUES (?)
            ON CONFLICT DO NOTHING
            RETURNING id
        """.trimIndent()
        ).use { stmt ->
            stmt.setBytes(1, whirlpool)

            stmt.executeQuery().use { rows ->
                if (rows.next()) {
                    rows.getLong(1)
                }
            }
        }

        connection.prepareStatement(
            """
            SELECT id
            FROM caches
            WHERE whirlpool = ?
        """.trimIndent()
        ).use { stmt ->
            stmt.setBytes(1, whirlpool)

            stmt.executeQuery().use { rows ->
                check(rows.next())
                return rows.getLong(1)
            }
        }
    }

    // TODO(gpe): skip most of this function if we encounter a conflict?
    private fun addIndex(connection: Connection, cacheId: Long, index: Index) {
        val containerId = ContainerImporter.addContainer(connection, index)

        connection.prepareStatement(
            """
            INSERT INTO indexes (container_id, version)
            VALUES (?, ?)
            ON CONFLICT DO NOTHING
        """.trimIndent()
        ).use { stmt ->
            stmt.setLong(1, containerId)
            stmt.setInt(2, index.index.version)
            stmt.execute()
        }

        connection.prepareStatement(
            """
            INSERT INTO cache_indexes (cache_id, archive_id, container_id)
            VALUES (?, ?, ?)
            ON CONFLICT DO NOTHING
        """.trimIndent()
        ).use { stmt ->
            stmt.setLong(1, cacheId)
            stmt.setInt(2, index.archive)
            stmt.setLong(3, containerId)
            stmt.execute()
        }

        connection.prepareStatement(
            """
            INSERT INTO index_groups (container_id, group_id, crc32, whirlpool, version, name_hash)
            VALUES (?, ?, ?, ?, ?, ?)
            ON CONFLICT DO NOTHING
        """.trimIndent()
        ).use { stmt ->
            for (group in index.index) {
                stmt.setLong(1, containerId)
                stmt.setInt(2, group.id)
                stmt.setInt(3, group.checksum)
                stmt.setBytes(4, group.digest)
                stmt.setInt(5, group.version)

                if (index.index.hasNames) {
                    stmt.setInt(6, group.nameHash)
                } else {
                    stmt.setNull(6, Types.INTEGER)
                }

                stmt.addBatch()
            }

            stmt.executeBatch()
        }

        connection.prepareStatement(
            """
            INSERT INTO index_files (container_id, group_id, file_id, name_hash)
            VALUES (?, ?, ?, ?)
            ON CONFLICT DO NOTHING
        """.trimIndent()
        ).use { stmt ->
            for (group in index.index) {
                for (file in group) {
                    stmt.setLong(1, containerId)
                    stmt.setInt(2, group.id)
                    stmt.setInt(3, file.id)

                    if (index.index.hasNames) {
                        stmt.setInt(4, file.nameHash)
                    } else {
                        stmt.setNull(4, Types.INTEGER)
                    }

                    stmt.addBatch()
                }
            }

            stmt.executeBatch()
        }
    }

    private companion object {
        private const val BATCH_SIZE = 1024
    }
}