Line data    Source code

       1             : // Copyright 2019 The LevelDB-Go and Pebble Authors. All rights reserved. Use
       2             : // of this source code is governed by a BSD-style license that can be found in
       3             : // the LICENSE file.
       4             : 
       5             : package sstable
       6             : 
       7             : import (
       8             :         "github.com/cockroachdb/pebble/internal/base"
       9             :         "github.com/cockroachdb/pebble/internal/cache"
      10             : )
      11             : 
      12             : // Compression is the per-block compression algorithm to use.
      13             : type Compression int
      14             : 
      15             : // The available compression types.
      16             : const (
      17             :         DefaultCompression Compression = iota
      18             :         NoCompression
      19             :         SnappyCompression
      20             :         ZstdCompression
      21             :         NCompression
      22             : )
      23             : 
      24             : var ignoredInternalProperties = map[string]struct{}{
      25             :         "rocksdb.column.family.id":             {},
      26             :         "rocksdb.fixed.key.length":             {},
      27             :         "rocksdb.index.key.is.user.key":        {},
      28             :         "rocksdb.index.value.is.delta.encoded": {},
      29             :         "rocksdb.oldest.key.time":              {},
      30             :         "rocksdb.creation.time":                {},
      31             :         "rocksdb.file.creation.time":           {},
      32             :         "rocksdb.format.version":               {},
      33             : }
      34             : 
      35           1 : func (c Compression) String() string {
      36           1 :         switch c {
      37           0 :         case DefaultCompression:
      38           0 :                 return "Default"
      39           1 :         case NoCompression:
      40           1 :                 return "NoCompression"
      41           1 :         case SnappyCompression:
      42           1 :                 return "Snappy"
      43           1 :         case ZstdCompression:
      44           1 :                 return "ZSTD"
      45           0 :         default:
      46           0 :                 return "Unknown"
      47             :         }
      48             : }
      49             : 
      50             : // FilterType exports the base.FilterType type.
      51             : type FilterType = base.FilterType
      52             : 
      53             : // Exported TableFilter constants.
      54             : const (
      55             :         TableFilter = base.TableFilter
      56             : )
      57             : 
      58             : // FilterWriter exports the base.FilterWriter type.
      59             : type FilterWriter = base.FilterWriter
      60             : 
      61             : // FilterPolicy exports the base.FilterPolicy type.
      62             : type FilterPolicy = base.FilterPolicy
      63             : 
      64             : // ReaderOptions holds the parameters needed for reading an sstable.
      65             : type ReaderOptions struct {
      66             :         // Cache is used to cache uncompressed blocks from sstables.
      67             :         //
      68             :         // The default cache size is a zero-size cache.
      69             :         Cache *cache.Cache
      70             : 
      71             :         // User properties specified in this map will not be added to sst.Properties.UserProperties.
      72             :         DeniedUserProperties map[string]struct{}
      73             : 
      74             :         // Comparer defines a total ordering over the space of []byte keys: a 'less
      75             :         // than' relationship. The same comparison algorithm must be used for reads
      76             :         // and writes over the lifetime of the DB.
      77             :         //
      78             :         // The default value uses the same ordering as bytes.Compare.
      79             :         Comparer *Comparer
      80             : 
      81             :         // Merge defines the Merge function in use for this keyspace.
      82             :         Merge base.Merge
      83             : 
      84             :         // Filters is a map from filter policy name to filter policy. It is used for
      85             :         // debugging tools which may be used on multiple databases configured with
      86             :         // different filter policies. It is not necessary to populate this filters
      87             :         // map during normal usage of a DB.
      88             :         Filters map[string]FilterPolicy
      89             : 
      90             :         // Merger defines the associative merge operation to use for merging values
      91             :         // written with {Batch,DB}.Merge. The MergerName is checked for consistency
      92             :         // with the value stored in the sstable when it was written.
      93             :         MergerName string
      94             : 
      95             :         // Logger is an optional logger and tracer.
      96             :         LoggerAndTracer base.LoggerAndTracer
      97             : }
      98             : 
      99           1 : func (o ReaderOptions) ensureDefaults() ReaderOptions {
     100           1 :         if o.Comparer == nil {
     101           1 :                 o.Comparer = base.DefaultComparer
     102           1 :         }
     103           1 :         if o.Merge == nil {
     104           1 :                 o.Merge = base.DefaultMerger.Merge
     105           1 :         }
     106           1 :         if o.MergerName == "" {
     107           1 :                 o.MergerName = base.DefaultMerger.Name
     108           1 :         }
     109           1 :         if o.LoggerAndTracer == nil {
     110           1 :                 o.LoggerAndTracer = base.NoopLoggerAndTracer{}
     111           1 :         }
     112           1 :         if o.DeniedUserProperties == nil {
     113           1 :                 o.DeniedUserProperties = ignoredInternalProperties
     114           1 :         }
     115           1 :         return o
     116             : }
     117             : 
     118             : // WriterOptions holds the parameters used to control building an sstable.
     119             : type WriterOptions struct {
     120             :         // BlockRestartInterval is the number of keys between restart points
     121             :         // for delta encoding of keys.
     122             :         //
     123             :         // The default value is 16.
     124             :         BlockRestartInterval int
     125             : 
     126             :         // BlockSize is the target uncompressed size in bytes of each table block.
     127             :         //
     128             :         // The default value is 4096.
     129             :         BlockSize int
     130             : 
     131             :         // BlockSizeThreshold finishes a block if the block size is larger than the
     132             :         // specified percentage of the target block size and adding the next entry
     133             :         // would cause the block to be larger than the target block size.
     134             :         //
     135             :         // The default value is 90.
     136             :         BlockSizeThreshold int
     137             : 
     138             :         // SizeClassAwareThreshold imposes a minimum block size restriction for blocks
     139             :         // to be flushed, that is computed as the percentage of the target block size.
     140             :         // Note that this threshold takes precedence over BlockSizeThreshold when
     141             :         // valid AllocatorSizeClasses are specified.
     142             :         //
     143             :         // The default value is 60.
     144             :         SizeClassAwareThreshold int
     145             : 
     146             :         // Cache is used to cache uncompressed blocks from sstables.
     147             :         //
     148             :         // The default is a nil cache.
     149             :         Cache *cache.Cache
     150             : 
     151             :         // Comparer defines a total ordering over the space of []byte keys: a 'less
     152             :         // than' relationship. The same comparison algorithm must be used for reads
     153             :         // and writes over the lifetime of the DB.
     154             :         //
     155             :         // The default value uses the same ordering as bytes.Compare.
     156             :         Comparer *Comparer
     157             : 
     158             :         // Compression defines the per-block compression to use.
     159             :         //
     160             :         // The default value (DefaultCompression) uses snappy compression.
     161             :         Compression Compression
     162             : 
     163             :         // FilterPolicy defines a filter algorithm (such as a Bloom filter) that can
     164             :         // reduce disk reads for Get calls.
     165             :         //
     166             :         // One such implementation is bloom.FilterPolicy(10) from the pebble/bloom
     167             :         // package.
     168             :         //
     169             :         // The default value means to use no filter.
     170             :         FilterPolicy FilterPolicy
     171             : 
     172             :         // FilterType defines whether an existing filter policy is applied at a
     173             :         // block-level or table-level. Block-level filters use less memory to create,
     174             :         // but are slower to access as a check for the key in the index must first be
     175             :         // performed to locate the filter block. A table-level filter will require
     176             :         // memory proportional to the number of keys in an sstable to create, but
     177             :         // avoids the index lookup when determining if a key is present. Table-level
     178             :         // filters should be preferred except under constrained memory situations.
     179             :         FilterType FilterType
     180             : 
     181             :         // IndexBlockSize is the target uncompressed size in bytes of each index
     182             :         // block. When the index block size is larger than this target, two-level
     183             :         // indexes are automatically enabled. Setting this option to a large value
     184             :         // (such as math.MaxInt32) disables the automatic creation of two-level
     185             :         // indexes.
     186             :         //
     187             :         // The default value is the value of BlockSize.
     188             :         IndexBlockSize int
     189             : 
     190             :         // Merger defines the associative merge operation to use for merging values
     191             :         // written with {Batch,DB}.Merge. The MergerName is checked for consistency
     192             :         // with the value stored in the sstable when it was written.
     193             :         MergerName string
     194             : 
     195             :         // TableFormat specifies the format version for writing sstables. The default
     196             :         // is TableFormatMinSupported.
     197             :         TableFormat TableFormat
     198             : 
     199             :         // IsStrictObsolete is only relevant for >= TableFormatPebblev4. See comment
     200             :         // in format.go. Must be false if format < TableFormatPebblev4.
     201             :         //
     202             :         // TODO(bilal): set this when writing shared ssts.
     203             :         IsStrictObsolete bool
     204             : 
     205             :         // WritingToLowestLevel is only relevant for >= TableFormatPebblev4. It is
     206             :         // used to set the obsolete bit on DEL/DELSIZED/SINGLEDEL if they are the
     207             :         // youngest for a userkey.
     208             :         WritingToLowestLevel bool
     209             : 
     210             :         // BlockPropertyCollectors is a list of BlockPropertyCollector creation
     211             :         // functions. A new BlockPropertyCollector is created for each sstable
     212             :         // built and lives for the lifetime of writing that table.
     213             :         BlockPropertyCollectors []func() BlockPropertyCollector
     214             : 
     215             :         // Checksum specifies which checksum to use.
     216             :         Checksum ChecksumType
     217             : 
     218             :         // Parallelism is used to indicate that the sstable Writer is allowed to
     219             :         // compress data blocks and write datablocks to disk in parallel with the
     220             :         // Writer client goroutine.
     221             :         Parallelism bool
     222             : 
     223             :         // ShortAttributeExtractor mirrors
     224             :         // Options.Experimental.ShortAttributeExtractor.
     225             :         ShortAttributeExtractor base.ShortAttributeExtractor
     226             : 
     227             :         // RequiredInPlaceValueBound mirrors
     228             :         // Options.Experimental.RequiredInPlaceValueBound.
     229             :         RequiredInPlaceValueBound UserKeyPrefixBound
     230             : 
     231             :         // DisableValueBlocks is only used for TableFormat >= TableFormatPebblev3,
     232             :         // and if set to true, does not write any values to value blocks. This is
     233             :         // only intended for cases where the in-memory buffering of all value blocks
     234             :         // while writing a sstable is too expensive and likely to cause an OOM. It
     235             :         // is never set to true by a Pebble DB, and can be set to true when some
     236             :         // external code is directly generating huge sstables using Pebble's
     237             :         // sstable.Writer (for example, CockroachDB backups can sometimes write
     238             :         // 750MB sstables -- see
     239             :         // https://github.com/cockroachdb/cockroach/issues/117113).
     240             :         DisableValueBlocks bool
     241             : 
     242             :         // AllocatorSizeClasses provides a sorted list containing the supported size
     243             :         // classes of the underlying memory allocator. This provides hints to the
     244             :         // writer's flushing policy to select block sizes that preemptively reduce
     245             :         // internal fragmentation when loaded into the block cache.
     246             :         AllocatorSizeClasses []int
     247             : }
     248             : 
     249           1 : func (o WriterOptions) ensureDefaults() WriterOptions {
     250           1 :         if o.BlockRestartInterval <= 0 {
     251           1 :                 o.BlockRestartInterval = base.DefaultBlockRestartInterval
     252           1 :         }
     253           1 :         if o.BlockSize <= 0 {
     254           1 :                 o.BlockSize = base.DefaultBlockSize
     255           1 :         }
     256           1 :         if o.BlockSizeThreshold <= 0 {
     257           1 :                 o.BlockSizeThreshold = base.DefaultBlockSizeThreshold
     258           1 :         }
     259           1 :         if o.SizeClassAwareThreshold <= 0 {
     260           1 :                 o.SizeClassAwareThreshold = base.SizeClassAwareBlockSizeThreshold
     261           1 :         }
     262           1 :         if o.Comparer == nil {
     263           1 :                 o.Comparer = base.DefaultComparer
     264           1 :         }
     265           1 :         if o.Compression <= DefaultCompression || o.Compression >= NCompression {
     266           1 :                 o.Compression = SnappyCompression
     267           1 :         }
     268           1 :         if o.IndexBlockSize <= 0 {
     269           1 :                 o.IndexBlockSize = o.BlockSize
     270           1 :         }
     271           1 :         if o.MergerName == "" {
     272           1 :                 o.MergerName = base.DefaultMerger.Name
     273           1 :         }
     274           1 :         if o.Checksum == ChecksumTypeNone {
     275           1 :                 o.Checksum = ChecksumTypeCRC32c
     276           1 :         }
     277             :         // By default, if the table format is not specified, fall back to using the
     278             :         // most compatible format that is supported by Pebble.
     279           1 :         if o.TableFormat == TableFormatUnspecified {
     280           1 :                 o.TableFormat = TableFormatMinSupported
     281           1 :         }
     282           1 :         return o
     283             : }