Class: Innodb::Page::Index

Inherits:
Innodb::Page show all
Extended by:
Forwardable
Defined in:
lib/innodb/page/index.rb

Direct Known Subclasses

IndexCompressed

Defined Under Namespace

Classes: FieldDescriptor, FsegHeader, PageHeader, RecordCursor, RecordHeader, SystemRecord, UserRecord

Constant Summary collapse

RECORD_NEXT_SIZE =

The size (in bytes) of the “next” pointer in each record header.

2
RECORD_REDUNDANT_BITS_SIZE =

The size (in bytes) of the bit-packed fields in each record header for “redundant” record format.

4
RECORD_REDUNDANT_OFF1_OFFSET_MASK =

Masks for 1-byte record end-offsets within “redundant” records.

0x7f
RECORD_REDUNDANT_OFF1_NULL_MASK = 
0x80
RECORD_REDUNDANT_OFF2_OFFSET_MASK =

Masks for 2-byte record end-offsets within “redundant” records.

0x3fff
RECORD_REDUNDANT_OFF2_NULL_MASK = 
0x8000
RECORD_REDUNDANT_OFF2_EXTERN_MASK = 
0x4000
RECORD_COMPACT_BITS_SIZE =

The size (in bytes) of the bit-packed fields in each record header for “compact” record format.

3
RECORD_MAX_N_SYSTEM_FIELDS =

Maximum number of fields.

3
RECORD_MAX_N_FIELDS = 
1024 - 1
RECORD_MAX_N_USER_FIELDS = 
RECORD_MAX_N_FIELDS - (RECORD_MAX_N_SYSTEM_FIELDS * 2)
PAGE_DIRECTION =

Page direction values possible in the page_header’s :direction field.

{
  1 => :left,           # Inserts have been in descending order.
  2 => :right,          # Inserts have been in ascending order.
  3 => :same_rec,       # Unused by InnoDB.
  4 => :same_page,      # Unused by InnoDB.
  5 => :no_direction,   # Inserts have been in random order.
}.freeze
RECORD_TYPES =

Record types used in the :type field of the record header.

{
  0 => :conventional,   # A normal user record in a leaf page.
  1 => :node_pointer,   # A node pointer in a non-leaf page.
  2 => :infimum,        # The system "infimum" record.
  3 => :supremum,       # The system "supremum" record.
}.freeze
PAGE_DIR_SLOT_SIZE =

The size (in bytes) of the record pointers in each page directory slot.

2
PAGE_DIR_SLOT_MIN_N_OWNED =

The minimum number of records “owned” by each record with an entry in the page directory.

4
PAGE_DIR_SLOT_MAX_N_OWNED =

The maximum number of records “owned” by each record with an entry in the page directory.

8

Constants inherited from Innodb::Page

PAGE_TYPE, PAGE_TYPE_BY_VALUE, UNDEFINED_PAGE_NUMBER

Instance Attribute Summary collapse

Attributes inherited from Innodb::Page

#space

Instance Method Summary collapse

Methods inherited from Innodb::Page

#checksum_crc32, #checksum_crc32?, #checksum_innodb, #checksum_innodb?, #checksum_invalid?, #checksum_type, #checksum_valid?, #corrupt?, #cursor, #default_page_size?, #each_page_body_byte_as_uint8, #each_page_header_byte_as_uint8, #extent_descriptor?, #fil_header, #fil_trailer, handle, #in_doublewrite_buffer?, #initialize, #inspect, #inspect_header_fields, maybe_undefined, #misplaced?, #misplaced_offset?, #misplaced_space?, #name, page_type_by_value, parse, #pos_fil_header, #pos_fil_trailer, #pos_page_body, #pos_partial_page_header, register_specialization, #size, #size_fil_header, #size_fil_trailer, #size_page_body, #size_partial_page_header, specialization_for, specialization_for?, #torn?, undefined?

Constructor Details

This class inherits a constructor from Innodb::Page

Instance Attribute Details

#record_describer ⇒ Object 



556
557
558
 
# File 'lib/innodb/page/index.rb', line 556

def record_describer
  @record_describer ||= make_record_describer
end

Instance Method Details

#binary_search_by_directory(dir, key) ⇒ Object

Search or a record within a single page using the page directory to limit the number of record comparisons required. Once the last page directory entry closest to but not greater than the key is found, fall back to linear search using linear_search_from_cursor to find the closest record whose key is not greater than the desired key. (If an exact match is desired, the returned record must be checked in the same way as the above linear_search_from_cursor function.)



884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
 
# File 'lib/innodb/page/index.rb', line 884

def binary_search_by_directory(dir, key)
  Innodb::Stats.increment :binary_search_by_directory

  return if dir.empty?

  # Split the directory at the mid-point (using integer math, so the division
  # is rounding down). Retrieve the record that sits at the mid-point.
  mid = ((dir.size - 1) / 2)
  rec = record(dir[mid])
  return unless rec

  if Innodb.debug?
    puts "binary_search_by_directory: page=%i, level=%i, dir.size=%i, dir[%i]=(%s)" % [
      offset,
      level,
      dir.size,
      mid,
      rec.key_string,
    ]
  end

  # The mid-point record was the infimum record, which is not comparable with
  # compare_key, so we need to just linear scan from here. If the mid-point
  # is the beginning of the page there can't be many records left to check
  # anyway.
  return linear_search_from_cursor(record_cursor(rec.next), key) if rec.header.type == :infimum

  # Compare the desired key to the mid-point record's key.
  case rec.compare_key(key)
  when 0
    # An exact match for the key was found. Return the record.
    Innodb::Stats.increment :binary_search_by_directory_exact_match
    rec
  when +1
    # The mid-point record's key is less than the desired key.
    if dir.size > 2
      # There are more entries remaining from the directory, recurse again
      # using binary search on the right half of the directory, which
      # represents values greater than or equal to the mid-point record's
      # key.
      Innodb::Stats.increment :binary_search_by_directory_recurse_right
      binary_search_by_directory(dir[mid...dir.size], key)
    else
      next_rec = record(dir[mid + 1])
      next_key = next_rec&.compare_key(key)
      if dir.size == 1 || next_key == -1 || next_key.zero?
        # This is the last entry remaining from the directory, or our key is
        # greater than rec and less than rec+1's key. Use linear search to
        # find the record starting at rec.
        Innodb::Stats.increment :binary_search_by_directory_linear_search
        linear_search_from_cursor(record_cursor(rec.offset), key)
      elsif next_key == +1
        Innodb::Stats.increment :binary_search_by_directory_linear_search
        linear_search_from_cursor(record_cursor(next_rec.offset), key)
      end
    end
  when -1
    # The mid-point record's key is greater than the desired key.
    if dir.size == 1
      # If this is the last entry remaining from the directory, we didn't
      # find anything workable.
      Innodb::Stats.increment :binary_search_by_directory_empty_result
      nil
    else
      # Recurse on the left half of the directory, which represents values
      # less than the mid-point record's key.
      Innodb::Stats.increment :binary_search_by_directory_recurse_left
      binary_search_by_directory(dir[0...mid], key)
    end
  end
end

#directory ⇒ Object

Return an array of row offsets for all entries in the page directory.



661
662
663
664
665
 
# File 'lib/innodb/page/index.rb', line 661

def directory
  @directory ||= cursor(pos_directory).backward.name("page_directory") do |c|
    directory_slots.times.map { |n| c.name("slot[#{n}]") { c.read_uint16 } }
  end
end

#directory_slot_for_record(this_record) ⇒ Object

Return the slot number for the page directory entry which “owns” the provided record. This will be either the record itself, or the nearest record with an entry in the directory and a value greater than the record.



682
683
684
685
686
687
688
689
690
691
692
693
694
695
 
# File 'lib/innodb/page/index.rb', line 682

def directory_slot_for_record(this_record)
  slot = record_directory_slot(this_record)
  return slot if slot

  search_cursor = record_cursor(this_record.next)
  raise "Could not position cursor" unless search_cursor

  while (rec = search_cursor.record)
    slot = record_directory_slot(rec)
    return slot if slot
  end

  record_directory_slot(supremum)
end

#directory_slots ⇒ Object

The number of directory slots in use.



264
265
266
 
# File 'lib/innodb/page/index.rb', line 264

def directory_slots
  page_header[:n_dir_slots]
end

#directory_space ⇒ Object

The amount of space consumed by the page directory.



269
270
271
 
# File 'lib/innodb/page/index.rb', line 269

def directory_space
  directory_slots * PAGE_DIR_SLOT_SIZE
end

#dump ⇒ Object

Dump the contents of a page for debugging purposes.



1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
 
# File 'lib/innodb/page/index.rb', line 1069

def dump
  super

  puts "page header:"
  pp page_header
  puts

  puts "fseg header:"
  pp fseg_header
  puts

  puts "sizes:"
  puts "  %-15s%5i" % ["header", header_space]
  puts "  %-15s%5i" % ["trailer", trailer_space]
  puts "  %-15s%5i" % ["directory", directory_space]
  puts "  %-15s%5i" % ["free", free_space]
  puts "  %-15s%5i" % ["used", used_space]
  puts "  %-15s%5i" % ["record", record_space]
  puts "  %-15s%5.2f" % ["per record", space_per_record]
  puts

  puts "page directory:"
  pp directory
  puts

  puts "system records:"
  pp infimum.record
  pp supremum.record
  puts

  if ibuf_index?
    puts "(records not dumped due to this being an insert buffer index)"
  elsif !record_describer
    puts "(records not dumped due to missing record describer or data dictionary)"
  else
    puts "garbage records:"
    each_garbage_record do |rec|
      pp rec.record
      puts
    end
    puts

    puts "records:"
    each_record do |rec|
      pp rec.record
      puts
    end
  end
  puts
end

#each_child_page ⇒ Object

Iterate through all child pages of a node (non-leaf) page, which are stored as records with the child page number as the last field in the record.



986
987
988
989
990
991
992
993
994
995
996
 
# File 'lib/innodb/page/index.rb', line 986

def each_child_page
  return if leaf?

  return enum_for(:each_child_page) unless block_given?

  each_record do |rec|
    yield rec.child_page_number, rec.key
  end

  nil
end

#each_directory_offset ⇒ Object 



697
698
699
700
701
702
703
 
# File 'lib/innodb/page/index.rb', line 697

def each_directory_offset
  return enum_for(:each_directory_offset) unless block_given?

  directory.each do |offset|
    yield offset unless [pos_infimum, pos_supremum].include?(offset)
  end
end

#each_directory_record ⇒ Object 



705
706
707
708
709
710
711
 
# File 'lib/innodb/page/index.rb', line 705

def each_directory_record
  return enum_for(:each_directory_record) unless block_given?

  each_directory_offset do |offset|
    yield record(offset)
  end
end

#each_garbage_record ⇒ Object

Iterate through all records in the garbage list.



970
971
972
973
974
975
976
977
978
979
980
981
 
# File 'lib/innodb/page/index.rb', line 970

def each_garbage_record
  return enum_for(:each_garbage_record) unless block_given?
  return if garbage_offset.zero?

  c = record_cursor(garbage_offset)

  while (rec = c.record)
    yield rec
  end

  nil
end

#each_record ⇒ Object

Iterate through all records.



957
958
959
960
961
962
963
964
965
966
967
 
# File 'lib/innodb/page/index.rb', line 957

def each_record
  return enum_for(:each_record) unless block_given?

  c = record_cursor(:min)

  while (rec = c.record)
    yield rec
  end

  nil
end

#each_region {|Region.new( offset: pos_index_header, length: size_index_header, name: :index_header, info: "Index Header" )| ... } ⇒ Object

Yields:



998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
 
# File 'lib/innodb/page/index.rb', line 998

def each_region(&block)
  return enum_for(:each_region) unless block_given?

  super

  yield Region.new(
    offset: pos_index_header,
    length: size_index_header,
    name: :index_header,
    info: "Index Header"
  )

  yield Region.new(
    offset: pos_fseg_header,
    length: size_fseg_header,
    name: :fseg_header,
    info: "File Segment Header"
  )

  yield Region.new(
    offset: pos_infimum - 5,
    length: size_mum_record + 5,
    name: :infimum,
    info: "Infimum"
  )

  yield Region.new(
    offset: pos_supremum - 5,
    length: size_mum_record + 5,
    name: :supremum,
    info: "Supremum"
  )

  directory_slots.times do |n|
    yield Region.new(
      offset: pos_directory - (n * 2),
      length: 2,
      name: :directory,
      info: "Page Directory"
    )
  end

  each_garbage_record do |record|
    yield Region.new(
      offset: record.offset - record.header.length,
      length: record.length + record.header.length,
      name: :garbage,
      info: "Garbage"
    )
  end

  each_record do |record|
    yield Region.new(
      offset: record.offset - record.header.length,
      length: record.header.length,
      name: :record_header,
      info: "Record Header"
    )

    yield Region.new(
      offset: record.offset,
      length: record.length || 1,
      name: :record_data,
      info: "Record Data"
    )
  end

  nil
end

#free_space ⇒ Object

Return the amount of free space in the page.



279
280
281
282
 
# File 'lib/innodb/page/index.rb', line 279

def free_space
  page_header[:garbage_size] +
    (size - size_fil_trailer - directory_space - page_header[:heap_top])
end

#fseg_header ⇒ Object

Return the “fseg” header.



346
347
348
349
350
351
352
353
 
# File 'lib/innodb/page/index.rb', line 346

def fseg_header
  @fseg_header ||= cursor(pos_fseg_header).name("fseg") do |c|
    FsegHeader.new(
      leaf: c.name("fseg[leaf]") { Innodb::FsegEntry.get_inode(@space, c) },
      internal: c.name("fseg[internal]") { Innodb::FsegEntry.get_inode(@space, c) }
    )
  end
end

#header_space ⇒ Object

The amount of space consumed by the page header.



257
258
259
260
261
 
# File 'lib/innodb/page/index.rb', line 257

def header_space
  # The end of the supremum system record is the beginning of the space
  # available for user records.
  pos_user_records
end

#ibuf_index? ⇒ Boolean

A helper to determine if an this page is part of an insert buffer index.

Returns:

  • (Boolean) 


341
342
343
 
# File 'lib/innodb/page/index.rb', line 341

def ibuf_index?
  index_id == Innodb::IbufIndex::INDEX_ID
end

#infimum ⇒ Object

Return the infimum record on a page.



539
540
541
 
# File 'lib/innodb/page/index.rb', line 539

def infimum
  @infimum ||= system_record(pos_infimum)
end

#leaf? ⇒ Boolean

A helper function to identify leaf index pages.

Returns:

  • (Boolean) 


336
337
338
 
# File 'lib/innodb/page/index.rb', line 336

def leaf?
  level.zero?
end

#linear_search_from_cursor(search_cursor, key) ⇒ Object

Search for a record within a single page, and return either a perfect match for the key, or the last record closest to they key but not greater than the key. (If an exact match is desired, compare_key must be used to check if the returned record matches. This makes the function useful for search in both leaf and non-leaf pages.)



834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
 
# File 'lib/innodb/page/index.rb', line 834

def linear_search_from_cursor(search_cursor, key)
  Innodb::Stats.increment :linear_search_from_cursor

  this_rec = search_cursor.record

  if Innodb.debug?
    puts "linear_search_from_cursor: page=%i, level=%i, start=(%s)" % [
      offset,
      level,
      this_rec && this_rec.key_string,
    ]
  end

  # Iterate through all records until finding either a matching record or
  # one whose key is greater than the desired key.
  while this_rec && (next_rec = search_cursor.record)
    Innodb::Stats.increment :linear_search_from_cursor_record_scans

    if Innodb.debug?
      puts "linear_search_from_cursor: page=%i, level=%i, current=(%s)" % [
        offset,
        level,
        this_rec.key_string,
      ]
    end

    # If we reach supremum, return the last non-system record we got.
    return this_rec if next_rec.header.type == :supremum

    return this_rec if this_rec.compare_key(key).negative?

    # The desired key is either an exact match for this_rec or is greater
    # than it but less than next_rec. If this is a non-leaf page, that
    # will mean that the record will fall on the leaf page this node
    # pointer record points to, if it exists at all.
    return this_rec if !this_rec.compare_key(key).negative? && next_rec.compare_key(key).negative?

    this_rec = next_rec
  end

  this_rec
end

#make_record_describer ⇒ Object 



548
549
550
551
552
553
554
 
# File 'lib/innodb/page/index.rb', line 548

def make_record_describer
  if space&.innodb_system&.data_dictionary&.found? && index_id && !ibuf_index?
    @record_describer = space.innodb_system.data_dictionary.record_describer_by_index_id(index_id)
  elsif space
    @record_describer = space.record_describer
  end
end

#make_record_description ⇒ Object

Return a set of field objects that describe the record.



561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
 
# File 'lib/innodb/page/index.rb', line 561

def make_record_description
  position = (0..RECORD_MAX_N_FIELDS).each
  description = record_describer.description
  fields = { type: description[:type], key: [], sys: [], row: [] }

  description[:key].each do |field|
    fields[:key] << Innodb::Field.new(position.next, field[:name], *field[:type])
  end

  # If this is a leaf page of the clustered index, read InnoDB's internal
  # fields, a transaction ID and roll pointer.
  if leaf? && fields[:type] == :clustered
    [["DB_TRX_ID", :TRX_ID], ["DB_ROLL_PTR", :ROLL_PTR]].each do |name, type|
      fields[:sys] << Innodb::Field.new(position.next, name, type, :NOT_NULL)
    end
  end

  # If this is a leaf page of the clustered index, or any page of a
  # secondary index, read the non-key fields.
  if (leaf? && fields[:type] == :clustered) || (fields[:type] == :secondary)
    description[:row].each do |field|
      fields[:row] << Innodb::Field.new(position.next, field[:name], *field[:type])
    end
  end

  fields
end

#max_record ⇒ Object

Return the maximum record on this page.



816
817
818
819
820
821
822
823
824
825
826
827
 
# File 'lib/innodb/page/index.rb', line 816

def max_record
  # Since the records are only singly-linked in the forward direction, in
  # order to do find the last record, we must create a cursor and walk
  # backwards one step.
  max_cursor = record_cursor(supremum.offset, :backward)
  raise "Could not position cursor" unless max_cursor

  # Note the deliberate use of prev_record rather than record; we want
  # to skip over supremum itself.
  max = max_cursor.prev_record
  max if max != infimum
end

#min_record ⇒ Object

Return the minimum record on this page.



810
811
812
813
 
# File 'lib/innodb/page/index.rb', line 810

def min_record
  min = record(infimum.next)
  min if min != supremum
end

#offset_directory_slot(offset) ⇒ Object

Return the slot number of the provided offset in the page directory, or nil if the offset is not present in the page directory.



669
670
671
 
# File 'lib/innodb/page/index.rb', line 669

def offset_directory_slot(offset)
  directory.index(offset)
end

#page_header ⇒ Object

Return the “index” header.



300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
 
# File 'lib/innodb/page/index.rb', line 300

def page_header
  @page_header ||= cursor(pos_index_header).name("index") do |c|
    index = PageHeader.new(
      n_dir_slots: c.name("n_dir_slots") { c.read_uint16 },
      heap_top: c.name("heap_top") { c.read_uint16 },
      n_heap_format: c.name("n_heap_format") { c.read_uint16 },
      garbage_offset: c.name("garbage_offset") { c.read_uint16 },
      garbage_size: c.name("garbage_size") { c.read_uint16 },
      last_insert_offset: c.name("last_insert_offset") { c.read_uint16 },
      direction: c.name("direction") { PAGE_DIRECTION[c.read_uint16] },
      n_direction: c.name("n_direction") { c.read_uint16 },
      n_recs: c.name("n_recs") { c.read_uint16 },
      max_trx_id: c.name("max_trx_id") { c.read_uint64 },
      level: c.name("level") { c.read_uint16 },
      index_id: c.name("index_id") { c.read_uint64 }
    )

    index.n_heap = index.n_heap_format & ((2**15) - 1)
    index.format = (index.n_heap_format & (1 << 15)).zero? ? :redundant : :compact

    index
  end
end

#pos_directory ⇒ Object

The position of the page directory, which starts at the “fil” trailer and grows backwards from there.



252
253
254
 
# File 'lib/innodb/page/index.rb', line 252

def pos_directory
  pos_fil_trailer
end

#pos_fseg_header ⇒ Object

Return the byte offset of the start of the “fseg” header, which immediately follows the “index” header.



178
179
180
 
# File 'lib/innodb/page/index.rb', line 178

def pos_fseg_header
  pos_index_header + size_index_header
end

#pos_index_header ⇒ Object

Return the byte offset of the start of the “index” page header, which immediately follows the “fil” header.



167
168
169
 
# File 'lib/innodb/page/index.rb', line 167

def pos_index_header
  pos_page_body
end

#pos_infimum ⇒ Object

Return the byte offset of the start of the “origin” of the infimum record, which is always the first record in the singly-linked record chain on any page, and represents a record with a “lower value than any possible user record”. The infimum record immediately follows the page header.



219
220
221
222
223
 
# File 'lib/innodb/page/index.rb', line 219

def pos_infimum
  pos_records +
    size_record_header +
    size_mum_record_header_additional
end

#pos_records ⇒ Object

Return the byte offset of the start of records within the page (the position immediately after the page header).



238
239
240
241
242
 
# File 'lib/innodb/page/index.rb', line 238

def pos_records
  size_fil_header +
    size_index_header +
    size_fseg_header
end

#pos_supremum ⇒ Object

Return the byte offset of the start of the “origin” of the supremum record, which is always the last record in the singly-linked record chain on any page, and represents a record with a “higher value than any possible user record”. The supremum record immediately follows the infimum record.



229
230
231
232
233
234
 
# File 'lib/innodb/page/index.rb', line 229

def pos_supremum
  pos_infimum +
    size_record_header +
    size_mum_record_header_additional +
    size_mum_record
end

#pos_user_records ⇒ Object

Return the byte offset of the start of the user records in a page, which immediately follows the supremum record.



246
247
248
 
# File 'lib/innodb/page/index.rb', line 246

def pos_user_records
  pos_supremum + size_mum_record
end

#record(offset) ⇒ Object

Parse and return a record at a given offset.



600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
 
# File 'lib/innodb/page/index.rb', line 600

def record(offset)
  return nil unless offset
  return infimum if offset == pos_infimum
  return supremum if offset == pos_supremum

  cursor(offset).forward.name("record[#{offset}]") do |c|
    # There is a header preceding the row itself, so back up and read it.
    header = c.peek { record_header(c) }

    this_record = UserRecord.new(
      format: page_header.format,
      offset: offset,
      header: header,
      next: header.next.zero? ? nil : header.next
    )

    if record_format
      this_record.type = record_format[:type]

      # Used to indicate whether a field is part of key/row/sys.
      # TODO: There's probably a better way to do this.
      fmap = %i[key row sys].each_with_object({}) do |k, h|
        this_record[k] = []
        record_format[k].each { |f| h[f.position] = k }
      end

      # Read the fields present in this record.
      record_fields.each do |f|
        p = fmap[f.position]
        c.name("#{p}[#{f.name}]") do
          this_record[p] << FieldDescriptor.new(
            name: f.name,
            type: f.data_type.name,
            value: f.value(c, this_record),
            extern: f.extern(c, this_record)
          )
        end
      end

      # If this is a node (non-leaf) page, it will have a child page number
      # (or "node pointer") stored as the last field.
      this_record.child_page_number = c.name("child_page_number") { c.read_uint32 } unless leaf?

      this_record.length = c.position - offset

      # Add system field accessors for convenience.
      this_record.sys.each do |f|
        case f[:name]
        when "DB_TRX_ID"
          this_record.transaction_id = f[:value]
        when "DB_ROLL_PTR"
          this_record.roll_pointer = f[:value]
        end
      end
    end

    Innodb::Record.new(self, this_record)
  end
end

#record_cursor(offset = :min, direction = :forward) ⇒ Object

Return a RecordCursor starting at offset.



799
800
801
 
# File 'lib/innodb/page/index.rb', line 799

def record_cursor(offset = :min, direction = :forward)
  RecordCursor.new(self, offset, direction)
end

#record_directory_slot(this_record) ⇒ Object

Return the slot number of the provided record in the page directory, or nil if the record is not present in the page directory.



675
676
677
 
# File 'lib/innodb/page/index.rb', line 675

def record_directory_slot(this_record)
  offset_directory_slot(this_record.offset)
end

#record_fields ⇒ Object

Returns the (ordered) set of fields that describe records in this page.



595
596
597
 
# File 'lib/innodb/page/index.rb', line 595

def record_fields
  record_format.values_at(:key, :sys, :row).flatten.sort_by(&:position) if record_format
end

#record_format ⇒ Object

Return (and cache) the record format provided by an external class.



590
591
592
 
# File 'lib/innodb/page/index.rb', line 590

def record_format
  @record_format ||= make_record_description if record_describer
end

#record_header(cursor) ⇒ Object

Return the header from a record.



356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
 
# File 'lib/innodb/page/index.rb', line 356

def record_header(cursor)
  origin = cursor.position
  header = RecordHeader.new
  cursor.backward.name("header") do |c|
    case page_header.format
    when :compact
      # The "next" pointer is a relative offset from the current record.
      header.next = c.name("next") { origin + c.read_sint16 }

      # Fields packed in a 16-bit integer (LSB first):
      #   3 bits for type
      #   13 bits for heap_number
      bits1 = c.name("bits1") { c.read_uint16 }
      header.type = RECORD_TYPES[bits1 & 0x07]
      header.heap_number = (bits1 & 0xfff8) >> 3
    when :redundant
      # The "next" pointer is an absolute offset within the page.
      header.next = c.name("next") { c.read_uint16 }

      # Fields packed in a 24-bit integer (LSB first):
      #   1 bit for offset_size (0 = 2 bytes, 1 = 1 byte)
      #   10 bits for n_fields
      #   13 bits for heap_number
      bits1 = c.name("bits1") { c.read_uint24 }
      header.offset_size = (bits1 & 1).zero? ? 2 : 1
      header.n_fields = (bits1 & (((1 << 10) - 1) << 1)) >> 1
      header.heap_number = (bits1 & (((1 << 13) - 1) << 11)) >> 11
    end

    # Fields packed in an 8-bit integer (LSB first):
    #   4 bits for n_owned
    #   4 bits for flags
    bits2 = c.name("bits2") { c.read_uint8 }
    header.n_owned = bits2 & 0x0f
    header.info_flags = (bits2 & 0xf0) >> 4

    case page_header.format
    when :compact
      record_header_compact_additional(header, cursor)
    when :redundant
      record_header_redundant_additional(header, cursor)
    end

    header.length = origin - cursor.position
  end

  header
end

#record_header_compact_additional(header, cursor) ⇒ Object

Read additional header information from a compact format record header.



406
407
408
409
410
411
412
413
414
415
416
417
418
419
 
# File 'lib/innodb/page/index.rb', line 406

def record_header_compact_additional(header, cursor)
  case header.type
  when :conventional, :node_pointer
    # The variable-length part of the record header contains a
    # bit vector indicating NULL fields and the length of each
    # non-NULL variable-length field.
    if record_format
      header.nulls = cursor.name("nulls") { record_header_compact_null_bitmap(cursor) }
      header.lengths, header.externs = cursor.name("lengths_and_externs") do
        record_header_compact_variable_lengths_and_externs(cursor, header.nulls)
      end
    end
  end
end

#record_header_compact_null_bitmap(cursor) ⇒ Object

Return an array indicating which fields are null.



422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
 
# File 'lib/innodb/page/index.rb', line 422

def record_header_compact_null_bitmap(cursor)
  fields = record_fields

  # The number of bits in the bitmap is the number of nullable fields.
  size = fields.count(&:nullable?)

  # There is no bitmap if there are no nullable fields.
  return [] unless size.positive?

  # TODO: This is really ugly.
  null_bit_array = cursor.read_bit_array(size).reverse!

  # For every nullable field, select the ones which are actually null.
  fields.select { |f| f.nullable? && (null_bit_array.shift == 1) }.map(&:name)
end

#record_header_compact_variable_lengths_and_externs(cursor, nulls) ⇒ Object

Return an array containing an array of the length of each variable-length field and an array indicating which fields are stored externally.



440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
 
# File 'lib/innodb/page/index.rb', line 440

def record_header_compact_variable_lengths_and_externs(cursor, nulls)
  fields = (record_format[:key] + record_format[:row])

  lengths = {}
  externs = []

  # For each non-NULL variable-length field, the record header contains
  # the length in one or two bytes.
  fields.each do |f|
    next if !f.variable? || nulls.include?(f.name)

    len = cursor.read_uint8
    ext = false

    # Two bytes are used only if the length exceeds 127 bytes and the
    # maximum length exceeds 255 bytes (or the field is a BLOB type).
    if len > 127 && (f.blob? || f.data_type.width > 255)
      ext = (0x40 & len) != 0
      len = ((len & 0x3f) << 8) + cursor.read_uint8
    end

    lengths[f.name] = len
    externs << f.name if ext
  end

  [lengths, externs]
end

#record_header_redundant_additional(header, cursor) ⇒ Object

Read additional header information from a redundant format record header.



469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
 
# File 'lib/innodb/page/index.rb', line 469

def record_header_redundant_additional(header, cursor)
  lengths = []
  nulls = []
  externs = []

  field_offsets = record_header_redundant_field_end_offsets(header, cursor)

  this_field_offset = 0
  field_offsets.each do |n|
    case header.offset_size
    when 1
      next_field_offset = (n & RECORD_REDUNDANT_OFF1_OFFSET_MASK)
      lengths << (next_field_offset - this_field_offset)
      nulls   << ((n & RECORD_REDUNDANT_OFF1_NULL_MASK) != 0)
      externs << false
    when 2
      next_field_offset = (n & RECORD_REDUNDANT_OFF2_OFFSET_MASK)
      lengths << (next_field_offset - this_field_offset)
      nulls   << ((n & RECORD_REDUNDANT_OFF2_NULL_MASK) != 0)
      externs << ((n & RECORD_REDUNDANT_OFF2_EXTERN_MASK) != 0)
    end
    this_field_offset = next_field_offset
  end

  # If possible, refer to fields by name rather than position for
  # better formatting (i.e. pp).
  if record_format
    header.lengths = {}
    header.nulls = []
    header.externs = []

    record_fields.each do |f|
      header.lengths[f.name] = lengths[f.position]
      header.nulls << f.name if nulls[f.position]
      header.externs << f.name if externs[f.position]
    end
  else
    header.lengths = lengths
    header.nulls = nulls
    header.externs = externs
  end
end

#record_header_redundant_field_end_offsets(header, cursor) ⇒ Object

Read field end offsets from the provided cursor for each field as counted by n_fields.



514
515
516
517
518
 
# File 'lib/innodb/page/index.rb', line 514

def record_header_redundant_field_end_offsets(header, cursor)
  header.n_fields.times.map do |n|
    cursor.name("field_end_offset[#{n}]") { cursor.read_uint_by_size(header.offset_size) }
  end
end

#record_if_exists(offset) ⇒ Object 



803
804
805
806
807
 
# File 'lib/innodb/page/index.rb', line 803

def record_if_exists(offset)
  each_record do |rec|
    return rec if rec.offset == offset
  end
end

#record_space ⇒ Object

Return the amount of space occupied by records in the page.



290
291
292
 
# File 'lib/innodb/page/index.rb', line 290

def record_space
  used_space - header_space - directory_space - trailer_space
end

#root? ⇒ Boolean

A helper function to identify root index pages; they must be the only pages at their level.

Returns:

  • (Boolean) 


331
332
333
 
# File 'lib/innodb/page/index.rb', line 331

def root?
  prev.nil? && self.next.nil?
end

#size_fseg_header ⇒ Object

The size of the “fseg” header.



183
184
185
 
# File 'lib/innodb/page/index.rb', line 183

def size_fseg_header
  2 * Innodb::FsegEntry::SIZE
end

#size_index_header ⇒ Object

The size of the “index” header.



172
173
174
 
# File 'lib/innodb/page/index.rb', line 172

def size_index_header
  2 + 2 + 2 + 2 + 2 + 2 + 2 + 2 + 2 + 8 + 2 + 8
end

#size_mum_record ⇒ Object

The size of the data from the supremum or infimum records.



211
212
213
 
# File 'lib/innodb/page/index.rb', line 211

def size_mum_record
  8
end

#size_mum_record_header_additional ⇒ Object

The size of the additional data structures in the header of the system records, which is just 1 byte in redundant format to store the offset of the end of the field. This is needed specifically here since we need to be able to calculate the fixed positions of these system records.



201
202
203
204
205
206
207
208
 
# File 'lib/innodb/page/index.rb', line 201

def size_mum_record_header_additional
  case page_header[:format]
  when :compact
    0 # No additional data is stored in compact format.
  when :redundant
    1 # A 1-byte offset for 1 field is stored in redundant format.
  end
end

#size_record_header ⇒ Object

Return the size of the header for each record.



188
189
190
191
192
193
194
195
 
# File 'lib/innodb/page/index.rb', line 188

def size_record_header
  case page_header[:format]
  when :compact
    RECORD_NEXT_SIZE + RECORD_COMPACT_BITS_SIZE
  when :redundant
    RECORD_NEXT_SIZE + RECORD_REDUNDANT_BITS_SIZE
  end
end

#space_per_record ⇒ Object

A helper to calculate the amount of space consumed per record.



295
296
297
 
# File 'lib/innodb/page/index.rb', line 295

def space_per_record
  page_header.n_recs.positive? ? (record_space.to_f / page_header.n_recs) : 0
end

#supremum ⇒ Object

Return the supremum record on a page.



544
545
546
 
# File 'lib/innodb/page/index.rb', line 544

def supremum
  @supremum ||= system_record(pos_supremum)
end

#system_record(offset) ⇒ Object

Parse and return simple fixed-format system records, such as InnoDB’s internal infimum and supremum records.



522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
 
# File 'lib/innodb/page/index.rb', line 522

def system_record(offset)
  cursor(offset).name("record[#{offset}]") do |c|
    header = c.peek { record_header(c) }
    Innodb::Record.new(
      self,
      SystemRecord.new(
        offset: offset,
        header: header,
        next: header.next,
        data: c.name("data") { c.read_bytes(size_mum_record) },
        length: c.position - offset
      )
    )
  end
end

#trailer_space ⇒ Object

The amount of space consumed by the trailers in the page.



274
275
276
 
# File 'lib/innodb/page/index.rb', line 274

def trailer_space
  size_fil_trailer
end

#used_space ⇒ Object

Return the amount of used space in the page.



285
286
287
 
# File 'lib/innodb/page/index.rb', line 285

def used_space
  size - free_space
end