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             : // CompressionFromString returns an sstable.Compression from its
      51             : // string representation. Inverse of c.String() above.
      52           1 : func CompressionFromString(s string) Compression {
      53           1 :         switch s {
      54           0 :         case "Default":
      55           0 :                 return DefaultCompression
      56           1 :         case "NoCompression":
      57           1 :                 return NoCompression
      58           1 :         case "Snappy":
      59           1 :                 return SnappyCompression
      60           1 :         case "ZSTD":
      61           1 :                 return ZstdCompression
      62           1 :         default:
      63           1 :                 return DefaultCompression
      64             :         }
      65             : }
      66             : 
      67             : // FilterType exports the base.FilterType type.
      68             : type FilterType = base.FilterType
      69             : 
      70             : // Exported TableFilter constants.
      71             : const (
      72             :         TableFilter = base.TableFilter
      73             : )
      74             : 
      75             : // FilterWriter exports the base.FilterWriter type.
      76             : type FilterWriter = base.FilterWriter
      77             : 
      78             : // FilterPolicy exports the base.FilterPolicy type.
      79             : type FilterPolicy = base.FilterPolicy
      80             : 
      81             : // ReaderOptions holds the parameters needed for reading an sstable.
      82             : type ReaderOptions struct {
      83             :         // Cache is used to cache uncompressed blocks from sstables.
      84             :         //
      85             :         // The default cache size is a zero-size cache.
      86             :         Cache *cache.Cache
      87             : 
      88             :         // User properties specified in this map will not be added to sst.Properties.UserProperties.
      89             :         DeniedUserProperties map[string]struct{}
      90             : 
      91             :         // Comparer defines a total ordering over the space of []byte keys: a 'less
      92             :         // than' relationship. The same comparison algorithm must be used for reads
      93             :         // and writes over the lifetime of the DB.
      94             :         //
      95             :         // The default value uses the same ordering as bytes.Compare.
      96             :         Comparer *Comparer
      97             : 
      98             :         // Merge defines the Merge function in use for this keyspace.
      99             :         Merge base.Merge
     100             : 
     101             :         // Filters is a map from filter policy name to filter policy. Filters with
     102             :         // policies that are not in this map will be ignored.
     103             :         Filters map[string]FilterPolicy
     104             : 
     105             :         // Merger defines the associative merge operation to use for merging values
     106             :         // written with {Batch,DB}.Merge. The MergerName is checked for consistency
     107             :         // with the value stored in the sstable when it was written.
     108             :         MergerName string
     109             : 
     110             :         // Logger is an optional logger and tracer.
     111             :         LoggerAndTracer base.LoggerAndTracer
     112             : }
     113             : 
     114           1 : func (o ReaderOptions) ensureDefaults() ReaderOptions {
     115           1 :         if o.Comparer == nil {
     116           0 :                 o.Comparer = base.DefaultComparer
     117           0 :         }
     118           1 :         if o.Merge == nil {
     119           1 :                 o.Merge = base.DefaultMerger.Merge
     120           1 :         }
     121           1 :         if o.MergerName == "" {
     122           1 :                 o.MergerName = base.DefaultMerger.Name
     123           1 :         }
     124           1 :         if o.LoggerAndTracer == nil {
     125           1 :                 o.LoggerAndTracer = base.NoopLoggerAndTracer{}
     126           1 :         }
     127           1 :         if o.DeniedUserProperties == nil {
     128           1 :                 o.DeniedUserProperties = ignoredInternalProperties
     129           1 :         }
     130           1 :         return o
     131             : }
     132             : 
     133             : // WriterOptions holds the parameters used to control building an sstable.
     134             : type WriterOptions struct {
     135             :         // BlockRestartInterval is the number of keys between restart points
     136             :         // for delta encoding of keys.
     137             :         //
     138             :         // The default value is 16.
     139             :         BlockRestartInterval int
     140             : 
     141             :         // BlockSize is the target uncompressed size in bytes of each table block.
     142             :         //
     143             :         // The default value is 4096.
     144             :         BlockSize int
     145             : 
     146             :         // BlockSizeThreshold finishes a block if the block size is larger than the
     147             :         // specified percentage of the target block size and adding the next entry
     148             :         // would cause the block to be larger than the target block size.
     149             :         //
     150             :         // The default value is 90.
     151             :         BlockSizeThreshold int
     152             : 
     153             :         // SizeClassAwareThreshold imposes a minimum block size restriction for blocks
     154             :         // to be flushed, that is computed as the percentage of the target block size.
     155             :         // Note that this threshold takes precedence over BlockSizeThreshold when
     156             :         // valid AllocatorSizeClasses are specified.
     157             :         //
     158             :         // The default value is 60.
     159             :         SizeClassAwareThreshold int
     160             : 
     161             :         // Cache is used to cache uncompressed blocks from sstables.
     162             :         //
     163             :         // The default is a nil cache.
     164             :         Cache *cache.Cache
     165             : 
     166             :         // Comparer defines a total ordering over the space of []byte keys: a 'less
     167             :         // than' relationship. The same comparison algorithm must be used for reads
     168             :         // and writes over the lifetime of the DB.
     169             :         //
     170             :         // The default value uses the same ordering as bytes.Compare.
     171             :         Comparer *Comparer
     172             : 
     173             :         // Compression defines the per-block compression to use.
     174             :         //
     175             :         // The default value (DefaultCompression) uses snappy compression.
     176             :         Compression Compression
     177             : 
     178             :         // FilterPolicy defines a filter algorithm (such as a Bloom filter) that can
     179             :         // reduce disk reads for Get calls.
     180             :         //
     181             :         // One such implementation is bloom.FilterPolicy(10) from the pebble/bloom
     182             :         // package.
     183             :         //
     184             :         // The default value means to use no filter.
     185             :         FilterPolicy FilterPolicy
     186             : 
     187             :         // FilterType defines whether an existing filter policy is applied at a
     188             :         // block-level or table-level. Block-level filters use less memory to create,
     189             :         // but are slower to access as a check for the key in the index must first be
     190             :         // performed to locate the filter block. A table-level filter will require
     191             :         // memory proportional to the number of keys in an sstable to create, but
     192             :         // avoids the index lookup when determining if a key is present. Table-level
     193             :         // filters should be preferred except under constrained memory situations.
     194             :         FilterType FilterType
     195             : 
     196             :         // IndexBlockSize is the target uncompressed size in bytes of each index
     197             :         // block. When the index block size is larger than this target, two-level
     198             :         // indexes are automatically enabled. Setting this option to a large value
     199             :         // (such as math.MaxInt32) disables the automatic creation of two-level
     200             :         // indexes.
     201             :         //
     202             :         // The default value is the value of BlockSize.
     203             :         IndexBlockSize int
     204             : 
     205             :         // Merger defines the associative merge operation to use for merging values
     206             :         // written with {Batch,DB}.Merge. The MergerName is checked for consistency
     207             :         // with the value stored in the sstable when it was written.
     208             :         MergerName string
     209             : 
     210             :         // TableFormat specifies the format version for writing sstables. The default
     211             :         // is TableFormatMinSupported.
     212             :         TableFormat TableFormat
     213             : 
     214             :         // IsStrictObsolete is only relevant for >= TableFormatPebblev4. See comment
     215             :         // in format.go. Must be false if format < TableFormatPebblev4.
     216             :         //
     217             :         // TODO(bilal): set this when writing shared ssts.
     218             :         IsStrictObsolete bool
     219             : 
     220             :         // WritingToLowestLevel is only relevant for >= TableFormatPebblev4. It is
     221             :         // used to set the obsolete bit on DEL/DELSIZED/SINGLEDEL if they are the
     222             :         // youngest for a userkey.
     223             :         WritingToLowestLevel bool
     224             : 
     225             :         // BlockPropertyCollectors is a list of BlockPropertyCollector creation
     226             :         // functions. A new BlockPropertyCollector is created for each sstable
     227             :         // built and lives for the lifetime of writing that table.
     228             :         BlockPropertyCollectors []func() BlockPropertyCollector
     229             : 
     230             :         // Checksum specifies which checksum to use.
     231             :         Checksum ChecksumType
     232             : 
     233             :         // Parallelism is used to indicate that the sstable Writer is allowed to
     234             :         // compress data blocks and write datablocks to disk in parallel with the
     235             :         // Writer client goroutine.
     236             :         Parallelism bool
     237             : 
     238             :         // ShortAttributeExtractor mirrors
     239             :         // Options.Experimental.ShortAttributeExtractor.
     240             :         ShortAttributeExtractor base.ShortAttributeExtractor
     241             : 
     242             :         // RequiredInPlaceValueBound mirrors
     243             :         // Options.Experimental.RequiredInPlaceValueBound.
     244             :         RequiredInPlaceValueBound UserKeyPrefixBound
     245             : 
     246             :         // DisableValueBlocks is only used for TableFormat >= TableFormatPebblev3,
     247             :         // and if set to true, does not write any values to value blocks. This is
     248             :         // only intended for cases where the in-memory buffering of all value blocks
     249             :         // while writing a sstable is too expensive and likely to cause an OOM. It
     250             :         // is never set to true by a Pebble DB, and can be set to true when some
     251             :         // external code is directly generating huge sstables using Pebble's
     252             :         // sstable.Writer (for example, CockroachDB backups can sometimes write
     253             :         // 750MB sstables -- see
     254             :         // https://github.com/cockroachdb/cockroach/issues/117113).
     255             :         DisableValueBlocks bool
     256             : 
     257             :         // AllocatorSizeClasses provides a sorted list containing the supported size
     258             :         // classes of the underlying memory allocator. This provides hints to the
     259             :         // writer's flushing policy to select block sizes that preemptively reduce
     260             :         // internal fragmentation when loaded into the block cache.
     261             :         AllocatorSizeClasses []int
     262             : }
     263             : 
     264           1 : func (o WriterOptions) ensureDefaults() WriterOptions {
     265           1 :         if o.BlockRestartInterval <= 0 {
     266           0 :                 o.BlockRestartInterval = base.DefaultBlockRestartInterval
     267           0 :         }
     268           1 :         if o.BlockSize <= 0 {
     269           0 :                 o.BlockSize = base.DefaultBlockSize
     270           0 :         }
     271           1 :         if o.BlockSizeThreshold <= 0 {
     272           0 :                 o.BlockSizeThreshold = base.DefaultBlockSizeThreshold
     273           0 :         }
     274           1 :         if o.SizeClassAwareThreshold <= 0 {
     275           1 :                 o.SizeClassAwareThreshold = base.SizeClassAwareBlockSizeThreshold
     276           1 :         }
     277           1 :         if o.Comparer == nil {
     278           0 :                 o.Comparer = base.DefaultComparer
     279           0 :         }
     280           1 :         if o.Compression <= DefaultCompression || o.Compression >= NCompression {
     281           0 :                 o.Compression = SnappyCompression
     282           0 :         }
     283           1 :         if o.IndexBlockSize <= 0 {
     284           0 :                 o.IndexBlockSize = o.BlockSize
     285           0 :         }
     286           1 :         if o.MergerName == "" {
     287           0 :                 o.MergerName = base.DefaultMerger.Name
     288           0 :         }
     289           1 :         if o.Checksum == ChecksumTypeNone {
     290           1 :                 o.Checksum = ChecksumTypeCRC32c
     291           1 :         }
     292             :         // By default, if the table format is not specified, fall back to using the
     293             :         // most compatible format that is supported by Pebble.
     294           1 :         if o.TableFormat == TableFormatUnspecified {
     295           0 :                 o.TableFormat = TableFormatMinSupported
     296           0 :         }
     297           1 :         return o
     298             : }