table.go

     1  // Copyright 2024 The Go Authors. All rights reserved.
     2  // Use of this source code is governed by a BSD-style
     3  // license that can be found in the LICENSE file.
     4  
     5  // Package maps implements Go's builtin map type.
     6  package maps
     7  
     8  import (
     9  	"internal/abi"
    10  	"internal/goarch"
    11  	"unsafe"
    12  )
    13  
    14  // Maximum size of a table before it is split at the directory level.
    15  //
    16  // TODO: Completely made up value. This should be tuned for performance vs grow
    17  // latency.
    18  // TODO: This should likely be based on byte size, as copying costs will
    19  // dominate grow latency for large objects.
    20  const maxTableCapacity = 1024
    21  
    22  // Ensure the max capacity fits in uint16, used for capacity and growthLeft
    23  // below.
    24  var _ = uint16(maxTableCapacity)
    25  
    26  // table is a Swiss table hash table structure.
    27  //
    28  // Each table is a complete hash table implementation.
    29  //
    30  // Map uses one or more tables to store entries. Extendible hashing (hash
    31  // prefix) is used to select the table to use for a specific key. Using
    32  // multiple tables enables incremental growth by growing only one table at a
    33  // time.
    34  type table struct {
    35  	// The number of filled slots (i.e. the number of elements in the table).
    36  	used uint16
    37  
    38  	// The total number of slots (always 2^N). Equal to
    39  	// `(groups.lengthMask+1)*abi.SwissMapGroupSlots`.
    40  	capacity uint16
    41  
    42  	// The number of slots we can still fill without needing to rehash.
    43  	//
    44  	// We rehash when used + tombstones > loadFactor*capacity, including
    45  	// tombstones so the table doesn't overfill with tombstones. This field
    46  	// counts down remaining empty slots before the next rehash.
    47  	growthLeft uint16
    48  
    49  	// The number of bits used by directory lookups above this table. Note
    50  	// that this may be less then globalDepth, if the directory has grown
    51  	// but this table has not yet been split.
    52  	localDepth uint8
    53  
    54  	// Index of this table in the Map directory. This is the index of the
    55  	// _first_ location in the directory. The table may occur in multiple
    56  	// sequential indicies.
    57  	//
    58  	// index is -1 if the table is stale (no longer installed in the
    59  	// directory).
    60  	index int
    61  
    62  	// groups is an array of slot groups. Each group holds abi.SwissMapGroupSlots
    63  	// key/elem slots and their control bytes. A table has a fixed size
    64  	// groups array. The table is replaced (in rehash) when more space is
    65  	// required.
    66  	//
    67  	// TODO(prattmic): keys and elements are interleaved to maximize
    68  	// locality, but it comes at the expense of wasted space for some types
    69  	// (consider uint8 key, uint64 element). Consider placing all keys
    70  	// together in these cases to save space.
    71  	groups groupsReference
    72  }
    73  
    74  func newTable(typ *abi.SwissMapType, capacity uint64, index int, localDepth uint8) *table {
    75  	if capacity < abi.SwissMapGroupSlots {
    76  		capacity = abi.SwissMapGroupSlots
    77  	}
    78  
    79  	t := &table{
    80  		index:      index,
    81  		localDepth: localDepth,
    82  	}
    83  
    84  	if capacity > maxTableCapacity {
    85  		panic("initial table capacity too large")
    86  	}
    87  
    88  	// N.B. group count must be a power of two for probeSeq to visit every
    89  	// group.
    90  	capacity, overflow := alignUpPow2(capacity)
    91  	if overflow {
    92  		panic("rounded-up capacity overflows uint64")
    93  	}
    94  
    95  	t.reset(typ, uint16(capacity))
    96  
    97  	return t
    98  }
    99  
   100  // reset resets the table with new, empty groups with the specified new total
   101  // capacity.
   102  func (t *table) reset(typ *abi.SwissMapType, capacity uint16) {
   103  	groupCount := uint64(capacity) / abi.SwissMapGroupSlots
   104  	t.groups = newGroups(typ, groupCount)
   105  	t.capacity = capacity
   106  	t.growthLeft = t.maxGrowthLeft()
   107  
   108  	for i := uint64(0); i <= t.groups.lengthMask; i++ {
   109  		g := t.groups.group(typ, i)
   110  		g.ctrls().setEmpty()
   111  	}
   112  }
   113  
   114  // maxGrowthLeft is the number of inserts we can do before
   115  // resizing, starting from an empty table.
   116  func (t *table) maxGrowthLeft() uint16 {
   117  	if t.capacity == 0 {
   118  		// No real reason to support zero capacity table, since an
   119  		// empty Map simply won't have a table.
   120  		panic("table must have positive capacity")
   121  	} else if t.capacity <= abi.SwissMapGroupSlots {
   122  		// If the map fits in a single group then we're able to fill all of
   123  		// the slots except 1 (an empty slot is needed to terminate find
   124  		// operations).
   125  		//
   126  		// TODO(go.dev/issue/54766): With a special case in probing for
   127  		// single-group tables, we could fill all slots.
   128  		return t.capacity - 1
   129  	} else {
   130  		if t.capacity*maxAvgGroupLoad < t.capacity {
   131  			// TODO(prattmic): Do something cleaner.
   132  			panic("overflow")
   133  		}
   134  		return (t.capacity * maxAvgGroupLoad) / abi.SwissMapGroupSlots
   135  	}
   136  
   137  }
   138  
   139  func (t *table) Used() uint64 {
   140  	return uint64(t.used)
   141  }
   142  
   143  // Get performs a lookup of the key that key points to. It returns a pointer to
   144  // the element, or false if the key doesn't exist.
   145  func (t *table) Get(typ *abi.SwissMapType, m *Map, key unsafe.Pointer) (unsafe.Pointer, bool) {
   146  	// TODO(prattmic): We could avoid hashing in a variety of special
   147  	// cases.
   148  	//
   149  	// - One entry maps could just directly compare the single entry
   150  	//   without hashing.
   151  	// - String keys could do quick checks of a few bytes before hashing.
   152  	hash := typ.Hasher(key, m.seed)
   153  	_, elem, ok := t.getWithKey(typ, hash, key)
   154  	return elem, ok
   155  }
   156  
   157  // getWithKey performs a lookup of key, returning a pointer to the version of
   158  // the key in the map in addition to the element.
   159  //
   160  // This is relevant when multiple different key values compare equal (e.g.,
   161  // +0.0 and -0.0). When a grow occurs during iteration, iteration perform a
   162  // lookup of keys from the old group in the new group in order to correctly
   163  // expose updated elements. For NeedsKeyUpdate keys, iteration also must return
   164  // the new key value, not the old key value.
   165  // hash must be the hash of the key.
   166  func (t *table) getWithKey(typ *abi.SwissMapType, hash uintptr, key unsafe.Pointer) (unsafe.Pointer, unsafe.Pointer, bool) {
   167  	// To find the location of a key in the table, we compute hash(key). From
   168  	// h1(hash(key)) and the capacity, we construct a probeSeq that visits
   169  	// every group of slots in some interesting order. See [probeSeq].
   170  	//
   171  	// We walk through these indices. At each index, we select the entire
   172  	// group starting with that index and extract potential candidates:
   173  	// occupied slots with a control byte equal to h2(hash(key)). The key
   174  	// at candidate slot i is compared with key; if key == g.slot(i).key
   175  	// we are done and return the slot; if there is an empty slot in the
   176  	// group, we stop and return an error; otherwise we continue to the
   177  	// next probe index. Tombstones (ctrlDeleted) effectively behave like
   178  	// full slots that never match the value we're looking for.
   179  	//
   180  	// The h2 bits ensure when we compare a key we are likely to have
   181  	// actually found the object. That is, the chance is low that keys
   182  	// compare false. Thus, when we search for an object, we are unlikely
   183  	// to call Equal many times. This likelihood can be analyzed as follows
   184  	// (assuming that h2 is a random enough hash function).
   185  	//
   186  	// Let's assume that there are k "wrong" objects that must be examined
   187  	// in a probe sequence. For example, when doing a find on an object
   188  	// that is in the table, k is the number of objects between the start
   189  	// of the probe sequence and the final found object (not including the
   190  	// final found object). The expected number of objects with an h2 match
   191  	// is then k/128. Measurements and analysis indicate that even at high
   192  	// load factors, k is less than 32, meaning that the number of false
   193  	// positive comparisons we must perform is less than 1/8 per find.
   194  	seq := makeProbeSeq(h1(hash), t.groups.lengthMask)
   195  	for ; ; seq = seq.next() {
   196  		g := t.groups.group(typ, seq.offset)
   197  
   198  		match := g.ctrls().matchH2(h2(hash))
   199  
   200  		for match != 0 {
   201  			i := match.first()
   202  
   203  			slotKey := g.key(typ, i)
   204  			if typ.IndirectKey() {
   205  				slotKey = *((*unsafe.Pointer)(slotKey))
   206  			}
   207  			if typ.Key.Equal(key, slotKey) {
   208  				slotElem := g.elem(typ, i)
   209  				if typ.IndirectElem() {
   210  					slotElem = *((*unsafe.Pointer)(slotElem))
   211  				}
   212  				return slotKey, slotElem, true
   213  			}
   214  			match = match.removeFirst()
   215  		}
   216  
   217  		match = g.ctrls().matchEmpty()
   218  		if match != 0 {
   219  			// Finding an empty slot means we've reached the end of
   220  			// the probe sequence.
   221  			return nil, nil, false
   222  		}
   223  	}
   224  }
   225  
   226  func (t *table) getWithoutKey(typ *abi.SwissMapType, hash uintptr, key unsafe.Pointer) (unsafe.Pointer, bool) {
   227  	seq := makeProbeSeq(h1(hash), t.groups.lengthMask)
   228  	for ; ; seq = seq.next() {
   229  		g := t.groups.group(typ, seq.offset)
   230  
   231  		match := g.ctrls().matchH2(h2(hash))
   232  
   233  		for match != 0 {
   234  			i := match.first()
   235  
   236  			slotKey := g.key(typ, i)
   237  			if typ.IndirectKey() {
   238  				slotKey = *((*unsafe.Pointer)(slotKey))
   239  			}
   240  			if typ.Key.Equal(key, slotKey) {
   241  				slotElem := g.elem(typ, i)
   242  				if typ.IndirectElem() {
   243  					slotElem = *((*unsafe.Pointer)(slotElem))
   244  				}
   245  				return slotElem, true
   246  			}
   247  			match = match.removeFirst()
   248  		}
   249  
   250  		match = g.ctrls().matchEmpty()
   251  		if match != 0 {
   252  			// Finding an empty slot means we've reached the end of
   253  			// the probe sequence.
   254  			return nil, false
   255  		}
   256  	}
   257  }
   258  
   259  // PutSlot returns a pointer to the element slot where an inserted element
   260  // should be written, and ok if it returned a valid slot.
   261  //
   262  // PutSlot returns ok false if the table was split and the Map needs to find
   263  // the new table.
   264  //
   265  // hash must be the hash of key.
   266  func (t *table) PutSlot(typ *abi.SwissMapType, m *Map, hash uintptr, key unsafe.Pointer) (unsafe.Pointer, bool) {
   267  	seq := makeProbeSeq(h1(hash), t.groups.lengthMask)
   268  
   269  	// As we look for a match, keep track of the first deleted slot we
   270  	// find, which we'll use to insert the new entry if necessary.
   271  	var firstDeletedGroup groupReference
   272  	var firstDeletedSlot uintptr
   273  
   274  	for ; ; seq = seq.next() {
   275  		g := t.groups.group(typ, seq.offset)
   276  		match := g.ctrls().matchH2(h2(hash))
   277  
   278  		// Look for an existing slot containing this key.
   279  		for match != 0 {
   280  			i := match.first()
   281  
   282  			slotKey := g.key(typ, i)
   283  			if typ.IndirectKey() {
   284  				slotKey = *((*unsafe.Pointer)(slotKey))
   285  			}
   286  			if typ.Key.Equal(key, slotKey) {
   287  				if typ.NeedKeyUpdate() {
   288  					typedmemmove(typ.Key, slotKey, key)
   289  				}
   290  
   291  				slotElem := g.elem(typ, i)
   292  				if typ.IndirectElem() {
   293  					slotElem = *((*unsafe.Pointer)(slotElem))
   294  				}
   295  
   296  				t.checkInvariants(typ, m)
   297  				return slotElem, true
   298  			}
   299  			match = match.removeFirst()
   300  		}
   301  
   302  		// No existing slot for this key in this group. Is this the end
   303  		// of the probe sequence?
   304  		match = g.ctrls().matchEmptyOrDeleted()
   305  		if match == 0 {
   306  			continue // nothing but filled slots. Keep probing.
   307  		}
   308  		i := match.first()
   309  		if g.ctrls().get(i) == ctrlDeleted {
   310  			// There are some deleted slots. Remember
   311  			// the first one, and keep probing.
   312  			if firstDeletedGroup.data == nil {
   313  				firstDeletedGroup = g
   314  				firstDeletedSlot = i
   315  			}
   316  			continue
   317  		}
   318  		// We've found an empty slot, which means we've reached the end of
   319  		// the probe sequence.
   320  
   321  		// If we found a deleted slot along the way, we can
   322  		// replace it without consuming growthLeft.
   323  		if firstDeletedGroup.data != nil {
   324  			g = firstDeletedGroup
   325  			i = firstDeletedSlot
   326  			t.growthLeft++ // will be decremented below to become a no-op.
   327  		}
   328  
   329  		// If we have no space left, first try to remove some tombstones.
   330  		if t.growthLeft == 0 {
   331  			t.pruneTombstones(typ, m)
   332  		}
   333  
   334  		// If there is room left to grow, just insert the new entry.
   335  		if t.growthLeft > 0 {
   336  			slotKey := g.key(typ, i)
   337  			if typ.IndirectKey() {
   338  				kmem := newobject(typ.Key)
   339  				*(*unsafe.Pointer)(slotKey) = kmem
   340  				slotKey = kmem
   341  			}
   342  			typedmemmove(typ.Key, slotKey, key)
   343  
   344  			slotElem := g.elem(typ, i)
   345  			if typ.IndirectElem() {
   346  				emem := newobject(typ.Elem)
   347  				*(*unsafe.Pointer)(slotElem) = emem
   348  				slotElem = emem
   349  			}
   350  
   351  			g.ctrls().set(i, ctrl(h2(hash)))
   352  			t.growthLeft--
   353  			t.used++
   354  			m.used++
   355  
   356  			t.checkInvariants(typ, m)
   357  			return slotElem, true
   358  		}
   359  
   360  		t.rehash(typ, m)
   361  		return nil, false
   362  	}
   363  }
   364  
   365  // uncheckedPutSlot inserts an entry known not to be in the table.
   366  // This is used for grow/split where we are making a new table from
   367  // entries in an existing table.
   368  //
   369  // Decrements growthLeft and increments used.
   370  //
   371  // Requires that the entry does not exist in the table, and that the table has
   372  // room for another element without rehashing.
   373  //
   374  // Requires that there are no deleted entries in the table.
   375  //
   376  // For indirect keys and/or elements, the key and elem pointers can be
   377  // put directly into the map, they do not need to be copied. This
   378  // requires the caller to ensure that the referenced memory never
   379  // changes (by sourcing those pointers from another indirect key/elem
   380  // map).
   381  func (t *table) uncheckedPutSlot(typ *abi.SwissMapType, hash uintptr, key, elem unsafe.Pointer) {
   382  	if t.growthLeft == 0 {
   383  		panic("invariant failed: growthLeft is unexpectedly 0")
   384  	}
   385  
   386  	// Given key and its hash hash(key), to insert it, we construct a
   387  	// probeSeq, and use it to find the first group with an unoccupied (empty
   388  	// or deleted) slot. We place the key/value into the first such slot in
   389  	// the group and mark it as full with key's H2.
   390  	seq := makeProbeSeq(h1(hash), t.groups.lengthMask)
   391  	for ; ; seq = seq.next() {
   392  		g := t.groups.group(typ, seq.offset)
   393  
   394  		match := g.ctrls().matchEmptyOrDeleted()
   395  		if match != 0 {
   396  			i := match.first()
   397  
   398  			slotKey := g.key(typ, i)
   399  			if typ.IndirectKey() {
   400  				*(*unsafe.Pointer)(slotKey) = key
   401  			} else {
   402  				typedmemmove(typ.Key, slotKey, key)
   403  			}
   404  
   405  			slotElem := g.elem(typ, i)
   406  			if typ.IndirectElem() {
   407  				*(*unsafe.Pointer)(slotElem) = elem
   408  			} else {
   409  				typedmemmove(typ.Elem, slotElem, elem)
   410  			}
   411  
   412  			t.growthLeft--
   413  			t.used++
   414  			g.ctrls().set(i, ctrl(h2(hash)))
   415  			return
   416  		}
   417  	}
   418  }
   419  
   420  // Delete returns true if it put a tombstone in t.
   421  func (t *table) Delete(typ *abi.SwissMapType, m *Map, hash uintptr, key unsafe.Pointer) bool {
   422  	seq := makeProbeSeq(h1(hash), t.groups.lengthMask)
   423  	for ; ; seq = seq.next() {
   424  		g := t.groups.group(typ, seq.offset)
   425  		match := g.ctrls().matchH2(h2(hash))
   426  
   427  		for match != 0 {
   428  			i := match.first()
   429  
   430  			slotKey := g.key(typ, i)
   431  			origSlotKey := slotKey
   432  			if typ.IndirectKey() {
   433  				slotKey = *((*unsafe.Pointer)(slotKey))
   434  			}
   435  
   436  			if typ.Key.Equal(key, slotKey) {
   437  				t.used--
   438  				m.used--
   439  
   440  				if typ.IndirectKey() {
   441  					// Clearing the pointer is sufficient.
   442  					*(*unsafe.Pointer)(origSlotKey) = nil
   443  				} else if typ.Key.Pointers() {
   444  					// Only bothing clear the key if there
   445  					// are pointers in it.
   446  					typedmemclr(typ.Key, slotKey)
   447  				}
   448  
   449  				slotElem := g.elem(typ, i)
   450  				if typ.IndirectElem() {
   451  					// Clearing the pointer is sufficient.
   452  					*(*unsafe.Pointer)(slotElem) = nil
   453  				} else {
   454  					// Unlike keys, always clear the elem (even if
   455  					// it contains no pointers), as compound
   456  					// assignment operations depend on cleared
   457  					// deleted values. See
   458  					// https://go.dev/issue/25936.
   459  					typedmemclr(typ.Elem, slotElem)
   460  				}
   461  
   462  				// Only a full group can appear in the middle
   463  				// of a probe sequence (a group with at least
   464  				// one empty slot terminates probing). Once a
   465  				// group becomes full, it stays full until
   466  				// rehashing/resizing. So if the group isn't
   467  				// full now, we can simply remove the element.
   468  				// Otherwise, we create a tombstone to mark the
   469  				// slot as deleted.
   470  				var tombstone bool
   471  				if g.ctrls().matchEmpty() != 0 {
   472  					g.ctrls().set(i, ctrlEmpty)
   473  					t.growthLeft++
   474  				} else {
   475  					g.ctrls().set(i, ctrlDeleted)
   476  					tombstone = true
   477  				}
   478  
   479  				t.checkInvariants(typ, m)
   480  				return tombstone
   481  			}
   482  			match = match.removeFirst()
   483  		}
   484  
   485  		match = g.ctrls().matchEmpty()
   486  		if match != 0 {
   487  			// Finding an empty slot means we've reached the end of
   488  			// the probe sequence.
   489  			return false
   490  		}
   491  	}
   492  }
   493  
   494  // pruneTombstones goes through the table and tries to remove
   495  // tombstones that are no longer needed. Best effort.
   496  // Note that it only removes tombstones, it does not move elements.
   497  // Moving elements would do a better job but is infeasbile due to
   498  // iterator semantics.
   499  //
   500  // Pruning should only succeed if it can remove O(n) tombstones.
   501  // It would be bad if we did O(n) work to find 1 tombstone to remove.
   502  // Then the next insert would spend another O(n) work to find 1 more
   503  // tombstone to remove, etc.
   504  //
   505  // We really need to remove O(n) tombstones so we can pay for the cost
   506  // of finding them. If we can't, then we need to grow (which is also O(n),
   507  // but guarantees O(n) subsequent inserts can happen in constant time).
   508  func (t *table) pruneTombstones(typ *abi.SwissMapType, m *Map) {
   509  	if t.tombstones()*10 < t.capacity { // 10% of capacity
   510  		// Not enough tombstones to be worth the effort.
   511  		return
   512  	}
   513  
   514  	// Bit set marking all the groups whose tombstones are needed.
   515  	var needed [(maxTableCapacity/abi.SwissMapGroupSlots + 31) / 32]uint32
   516  
   517  	// Trace the probe sequence of every full entry.
   518  	for i := uint64(0); i <= t.groups.lengthMask; i++ {
   519  		g := t.groups.group(typ, i)
   520  		match := g.ctrls().matchFull()
   521  		for match != 0 {
   522  			j := match.first()
   523  			match = match.removeFirst()
   524  			key := g.key(typ, j)
   525  			if typ.IndirectKey() {
   526  				key = *((*unsafe.Pointer)(key))
   527  			}
   528  			if !typ.Key.Equal(key, key) {
   529  				// Key not equal to itself. We never have to find these
   530  				// keys on lookup (only on iteration), so we can break
   531  				// their probe sequences at will.
   532  				continue
   533  			}
   534  			// Walk probe sequence for this key.
   535  			// Each tombstone group we need to walk past is marked required.
   536  			hash := typ.Hasher(key, m.seed)
   537  			for seq := makeProbeSeq(h1(hash), t.groups.lengthMask); ; seq = seq.next() {
   538  				if seq.offset == i {
   539  					break // reached group of element in probe sequence
   540  				}
   541  				g := t.groups.group(typ, seq.offset)
   542  				m := g.ctrls().matchEmptyOrDeleted()
   543  				if m != 0 { // must be deleted, not empty, as we haven't found our key yet
   544  					// Mark this group's tombstone as required.
   545  					needed[seq.offset/32] |= 1 << (seq.offset % 32)
   546  				}
   547  			}
   548  		}
   549  		if g.ctrls().matchEmpty() != 0 {
   550  			// Also mark non-tombstone-containing groups, so we don't try
   551  			// to remove tombstones from them below.
   552  			needed[i/32] |= 1 << (i % 32)
   553  		}
   554  	}
   555  
   556  	// First, see if we can remove enough tombstones to restore capacity.
   557  	// This function is O(n), so only remove tombstones if we can remove
   558  	// enough of them to justify the O(n) cost.
   559  	cnt := 0
   560  	for i := uint64(0); i <= t.groups.lengthMask; i++ {
   561  		if needed[i/32]>>(i%32)&1 != 0 {
   562  			continue
   563  		}
   564  		g := t.groups.group(typ, i)
   565  		m := g.ctrls().matchEmptyOrDeleted() // must be deleted
   566  		cnt += m.count()
   567  	}
   568  	if cnt*10 < int(t.capacity) { // Can we restore 10% of capacity?
   569  		return // don't bother removing tombstones. Caller will grow instead.
   570  	}
   571  
   572  	// Prune unneeded tombstones.
   573  	for i := uint64(0); i <= t.groups.lengthMask; i++ {
   574  		if needed[i/32]>>(i%32)&1 != 0 {
   575  			continue
   576  		}
   577  		g := t.groups.group(typ, i)
   578  		m := g.ctrls().matchEmptyOrDeleted() // must be deleted
   579  		for m != 0 {
   580  			k := m.first()
   581  			m = m.removeFirst()
   582  			g.ctrls().set(k, ctrlEmpty)
   583  			t.growthLeft++
   584  		}
   585  		// TODO: maybe we could convert all slots at once
   586  		// using some bitvector trickery.
   587  	}
   588  }
   589  
   590  // tombstones returns the number of deleted (tombstone) entries in the table. A
   591  // tombstone is a slot that has been deleted but is still considered occupied
   592  // so as not to violate the probing invariant.
   593  func (t *table) tombstones() uint16 {
   594  	return (t.capacity*maxAvgGroupLoad)/abi.SwissMapGroupSlots - t.used - t.growthLeft
   595  }
   596  
   597  // Clear deletes all entries from the map resulting in an empty map.
   598  func (t *table) Clear(typ *abi.SwissMapType) {
   599  	mgl := t.maxGrowthLeft()
   600  	if t.used == 0 && t.growthLeft == mgl { // no current entries and no tombstones
   601  		return
   602  	}
   603  	for i := uint64(0); i <= t.groups.lengthMask; i++ {
   604  		g := t.groups.group(typ, i)
   605  		if g.ctrls().matchFull() != 0 {
   606  			typedmemclr(typ.Group, g.data)
   607  		}
   608  		g.ctrls().setEmpty()
   609  	}
   610  	t.used = 0
   611  	t.growthLeft = mgl
   612  }
   613  
   614  type Iter struct {
   615  	key  unsafe.Pointer // Must be in first position.  Write nil to indicate iteration end (see cmd/compile/internal/walk/range.go).
   616  	elem unsafe.Pointer // Must be in second position (see cmd/compile/internal/walk/range.go).
   617  	typ  *abi.SwissMapType
   618  	m    *Map
   619  
   620  	// Randomize iteration order by starting iteration at a random slot
   621  	// offset. The offset into the directory uses a separate offset, as it
   622  	// must adjust when the directory grows.
   623  	entryOffset uint64
   624  	dirOffset   uint64
   625  
   626  	// Snapshot of Map.clearSeq at iteration initialization time. Used to
   627  	// detect clear during iteration.
   628  	clearSeq uint64
   629  
   630  	// Value of Map.globalDepth during the last call to Next. Used to
   631  	// detect directory grow during iteration.
   632  	globalDepth uint8
   633  
   634  	// dirIdx is the current directory index, prior to adjustment by
   635  	// dirOffset.
   636  	dirIdx int
   637  
   638  	// tab is the table at dirIdx during the previous call to Next.
   639  	tab *table
   640  
   641  	// group is the group at entryIdx during the previous call to Next.
   642  	group groupReference
   643  
   644  	// entryIdx is the current entry index, prior to adjustment by entryOffset.
   645  	// The lower 3 bits of the index are the slot index, and the upper bits
   646  	// are the group index.
   647  	entryIdx uint64
   648  }
   649  
   650  // Init initializes Iter for iteration.
   651  func (it *Iter) Init(typ *abi.SwissMapType, m *Map) {
   652  	it.typ = typ
   653  
   654  	if m == nil || m.used == 0 {
   655  		return
   656  	}
   657  
   658  	dirIdx := 0
   659  	var groupSmall groupReference
   660  	if m.dirLen <= 0 {
   661  		// Use dirIdx == -1 as sentinel for small maps.
   662  		dirIdx = -1
   663  		groupSmall.data = m.dirPtr
   664  	}
   665  
   666  	it.m = m
   667  	it.entryOffset = rand()
   668  	it.dirOffset = rand()
   669  	it.globalDepth = m.globalDepth
   670  	it.dirIdx = dirIdx
   671  	it.group = groupSmall
   672  	it.clearSeq = m.clearSeq
   673  }
   674  
   675  func (it *Iter) Initialized() bool {
   676  	return it.typ != nil
   677  }
   678  
   679  // Map returns the map this iterator is iterating over.
   680  func (it *Iter) Map() *Map {
   681  	return it.m
   682  }
   683  
   684  // Key returns a pointer to the current key. nil indicates end of iteration.
   685  //
   686  // Must not be called prior to Next.
   687  func (it *Iter) Key() unsafe.Pointer {
   688  	return it.key
   689  }
   690  
   691  // Key returns a pointer to the current element. nil indicates end of
   692  // iteration.
   693  //
   694  // Must not be called prior to Next.
   695  func (it *Iter) Elem() unsafe.Pointer {
   696  	return it.elem
   697  }
   698  
   699  func (it *Iter) nextDirIdx() {
   700  	// Skip other entries in the directory that refer to the same
   701  	// logical table. There are two cases of this:
   702  	//
   703  	// Consider this directory:
   704  	//
   705  	// - 0: *t1
   706  	// - 1: *t1
   707  	// - 2: *t2a
   708  	// - 3: *t2b
   709  	//
   710  	// At some point, the directory grew to accommodate a split of
   711  	// t2. t1 did not split, so entries 0 and 1 both point to t1.
   712  	// t2 did split, so the two halves were installed in entries 2
   713  	// and 3.
   714  	//
   715  	// If dirIdx is 0 and it.tab is t1, then we should skip past
   716  	// entry 1 to avoid repeating t1.
   717  	//
   718  	// If dirIdx is 2 and it.tab is t2 (pre-split), then we should
   719  	// skip past entry 3 because our pre-split t2 already covers
   720  	// all keys from t2a and t2b (except for new insertions, which
   721  	// iteration need not return).
   722  	//
   723  	// We can achieve both of these by using to difference between
   724  	// the directory and table depth to compute how many entries
   725  	// the table covers.
   726  	entries := 1 << (it.m.globalDepth - it.tab.localDepth)
   727  	it.dirIdx += entries
   728  	it.tab = nil
   729  	it.group = groupReference{}
   730  	it.entryIdx = 0
   731  }
   732  
   733  // Return the appropriate key/elem for key at slotIdx index within it.group, if
   734  // any.
   735  func (it *Iter) grownKeyElem(key unsafe.Pointer, slotIdx uintptr) (unsafe.Pointer, unsafe.Pointer, bool) {
   736  	newKey, newElem, ok := it.m.getWithKey(it.typ, key)
   737  	if !ok {
   738  		// Key has likely been deleted, and
   739  		// should be skipped.
   740  		//
   741  		// One exception is keys that don't
   742  		// compare equal to themselves (e.g.,
   743  		// NaN). These keys cannot be looked
   744  		// up, so getWithKey will fail even if
   745  		// the key exists.
   746  		//
   747  		// However, we are in luck because such
   748  		// keys cannot be updated and they
   749  		// cannot be deleted except with clear.
   750  		// Thus if no clear has occurred, the
   751  		// key/elem must still exist exactly as
   752  		// in the old groups, so we can return
   753  		// them from there.
   754  		//
   755  		// TODO(prattmic): Consider checking
   756  		// clearSeq early. If a clear occurred,
   757  		// Next could always return
   758  		// immediately, as iteration doesn't
   759  		// need to return anything added after
   760  		// clear.
   761  		if it.clearSeq == it.m.clearSeq && !it.typ.Key.Equal(key, key) {
   762  			elem := it.group.elem(it.typ, slotIdx)
   763  			if it.typ.IndirectElem() {
   764  				elem = *((*unsafe.Pointer)(elem))
   765  			}
   766  			return key, elem, true
   767  		}
   768  
   769  		// This entry doesn't exist anymore.
   770  		return nil, nil, false
   771  	}
   772  
   773  	return newKey, newElem, true
   774  }
   775  
   776  // Next proceeds to the next element in iteration, which can be accessed via
   777  // the Key and Elem methods.
   778  //
   779  // The table can be mutated during iteration, though there is no guarantee that
   780  // the mutations will be visible to the iteration.
   781  //
   782  // Init must be called prior to Next.
   783  func (it *Iter) Next() {
   784  	if it.m == nil {
   785  		// Map was empty at Iter.Init.
   786  		it.key = nil
   787  		it.elem = nil
   788  		return
   789  	}
   790  
   791  	if it.m.writing != 0 {
   792  		fatal("concurrent map iteration and map write")
   793  		return
   794  	}
   795  
   796  	if it.dirIdx < 0 {
   797  		// Map was small at Init.
   798  		for ; it.entryIdx < abi.SwissMapGroupSlots; it.entryIdx++ {
   799  			k := uintptr(it.entryIdx+it.entryOffset) % abi.SwissMapGroupSlots
   800  
   801  			if (it.group.ctrls().get(k) & ctrlEmpty) == ctrlEmpty {
   802  				// Empty or deleted.
   803  				continue
   804  			}
   805  
   806  			key := it.group.key(it.typ, k)
   807  			if it.typ.IndirectKey() {
   808  				key = *((*unsafe.Pointer)(key))
   809  			}
   810  
   811  			// As below, if we have grown to a full map since Init,
   812  			// we continue to use the old group to decide the keys
   813  			// to return, but must look them up again in the new
   814  			// tables.
   815  			grown := it.m.dirLen > 0
   816  			var elem unsafe.Pointer
   817  			if grown {
   818  				var ok bool
   819  				newKey, newElem, ok := it.m.getWithKey(it.typ, key)
   820  				if !ok {
   821  					// See comment below.
   822  					if it.clearSeq == it.m.clearSeq && !it.typ.Key.Equal(key, key) {
   823  						elem = it.group.elem(it.typ, k)
   824  						if it.typ.IndirectElem() {
   825  							elem = *((*unsafe.Pointer)(elem))
   826  						}
   827  					} else {
   828  						continue
   829  					}
   830  				} else {
   831  					key = newKey
   832  					elem = newElem
   833  				}
   834  			} else {
   835  				elem = it.group.elem(it.typ, k)
   836  				if it.typ.IndirectElem() {
   837  					elem = *((*unsafe.Pointer)(elem))
   838  				}
   839  			}
   840  
   841  			it.entryIdx++
   842  			it.key = key
   843  			it.elem = elem
   844  			return
   845  		}
   846  		it.key = nil
   847  		it.elem = nil
   848  		return
   849  	}
   850  
   851  	if it.globalDepth != it.m.globalDepth {
   852  		// Directory has grown since the last call to Next. Adjust our
   853  		// directory index.
   854  		//
   855  		// Consider:
   856  		//
   857  		// Before:
   858  		// - 0: *t1
   859  		// - 1: *t2  <- dirIdx
   860  		//
   861  		// After:
   862  		// - 0: *t1a (split)
   863  		// - 1: *t1b (split)
   864  		// - 2: *t2  <- dirIdx
   865  		// - 3: *t2
   866  		//
   867  		// That is, we want to double the current index when the
   868  		// directory size doubles (or quadruple when the directory size
   869  		// quadruples, etc).
   870  		//
   871  		// The actual (randomized) dirIdx is computed below as:
   872  		//
   873  		// dirIdx := (it.dirIdx + it.dirOffset) % it.m.dirLen
   874  		//
   875  		// Multiplication is associative across modulo operations,
   876  		// A * (B % C) = (A * B) % (A * C),
   877  		// provided that A is positive.
   878  		//
   879  		// Thus we can achieve this by adjusting it.dirIdx,
   880  		// it.dirOffset, and it.m.dirLen individually.
   881  		orders := it.m.globalDepth - it.globalDepth
   882  		it.dirIdx <<= orders
   883  		it.dirOffset <<= orders
   884  		// it.m.dirLen was already adjusted when the directory grew.
   885  
   886  		it.globalDepth = it.m.globalDepth
   887  	}
   888  
   889  	// Continue iteration until we find a full slot.
   890  	for ; it.dirIdx < it.m.dirLen; it.nextDirIdx() {
   891  		// Resolve the table.
   892  		if it.tab == nil {
   893  			dirIdx := int((uint64(it.dirIdx) + it.dirOffset) & uint64(it.m.dirLen-1))
   894  			newTab := it.m.directoryAt(uintptr(dirIdx))
   895  			if newTab.index != dirIdx {
   896  				// Normally we skip past all duplicates of the
   897  				// same entry in the table (see updates to
   898  				// it.dirIdx at the end of the loop below), so
   899  				// this case wouldn't occur.
   900  				//
   901  				// But on the very first call, we have a
   902  				// completely randomized dirIdx that may refer
   903  				// to a middle of a run of tables in the
   904  				// directory. Do a one-time adjustment of the
   905  				// offset to ensure we start at first index for
   906  				// newTable.
   907  				diff := dirIdx - newTab.index
   908  				it.dirOffset -= uint64(diff)
   909  				dirIdx = newTab.index
   910  			}
   911  			it.tab = newTab
   912  		}
   913  
   914  		// N.B. Use it.tab, not newTab. It is important to use the old
   915  		// table for key selection if the table has grown. See comment
   916  		// on grown below.
   917  
   918  		entryMask := uint64(it.tab.capacity) - 1
   919  		if it.entryIdx > entryMask {
   920  			// Continue to next table.
   921  			continue
   922  		}
   923  
   924  		// Fast path: skip matching and directly check if entryIdx is a
   925  		// full slot.
   926  		//
   927  		// In the slow path below, we perform an 8-slot match check to
   928  		// look for full slots within the group.
   929  		//
   930  		// However, with a max load factor of 7/8, each slot in a
   931  		// mostly full map has a high probability of being full. Thus
   932  		// it is cheaper to check a single slot than do a full control
   933  		// match.
   934  
   935  		entryIdx := (it.entryIdx + it.entryOffset) & entryMask
   936  		slotIdx := uintptr(entryIdx & (abi.SwissMapGroupSlots - 1))
   937  		if slotIdx == 0 || it.group.data == nil {
   938  			// Only compute the group (a) when we switch
   939  			// groups (slotIdx rolls over) and (b) on the
   940  			// first iteration in this table (slotIdx may
   941  			// not be zero due to entryOffset).
   942  			groupIdx := entryIdx >> abi.SwissMapGroupSlotsBits
   943  			it.group = it.tab.groups.group(it.typ, groupIdx)
   944  		}
   945  
   946  		if (it.group.ctrls().get(slotIdx) & ctrlEmpty) == 0 {
   947  			// Slot full.
   948  
   949  			key := it.group.key(it.typ, slotIdx)
   950  			if it.typ.IndirectKey() {
   951  				key = *((*unsafe.Pointer)(key))
   952  			}
   953  
   954  			grown := it.tab.index == -1
   955  			var elem unsafe.Pointer
   956  			if grown {
   957  				newKey, newElem, ok := it.grownKeyElem(key, slotIdx)
   958  				if !ok {
   959  					// This entry doesn't exist
   960  					// anymore. Continue to the
   961  					// next one.
   962  					goto next
   963  				} else {
   964  					key = newKey
   965  					elem = newElem
   966  				}
   967  			} else {
   968  				elem = it.group.elem(it.typ, slotIdx)
   969  				if it.typ.IndirectElem() {
   970  					elem = *((*unsafe.Pointer)(elem))
   971  				}
   972  			}
   973  
   974  			it.entryIdx++
   975  			it.key = key
   976  			it.elem = elem
   977  			return
   978  		}
   979  
   980  	next:
   981  		it.entryIdx++
   982  
   983  		// Slow path: use a match on the control word to jump ahead to
   984  		// the next full slot.
   985  		//
   986  		// This is highly effective for maps with particularly low load
   987  		// (e.g., map allocated with large hint but few insertions).
   988  		//
   989  		// For maps with medium load (e.g., 3-4 empty slots per group)
   990  		// it also tends to work pretty well. Since slots within a
   991  		// group are filled in order, then if there have been no
   992  		// deletions, a match will allow skipping past all empty slots
   993  		// at once.
   994  		//
   995  		// Note: it is tempting to cache the group match result in the
   996  		// iterator to use across Next calls. However because entries
   997  		// may be deleted between calls later calls would still need to
   998  		// double-check the control value.
   999  
  1000  		var groupMatch bitset
  1001  		for it.entryIdx <= entryMask {
  1002  			entryIdx := (it.entryIdx + it.entryOffset) & entryMask
  1003  			slotIdx := uintptr(entryIdx & (abi.SwissMapGroupSlots - 1))
  1004  
  1005  			if slotIdx == 0 || it.group.data == nil {
  1006  				// Only compute the group (a) when we switch
  1007  				// groups (slotIdx rolls over) and (b) on the
  1008  				// first iteration in this table (slotIdx may
  1009  				// not be zero due to entryOffset).
  1010  				groupIdx := entryIdx >> abi.SwissMapGroupSlotsBits
  1011  				it.group = it.tab.groups.group(it.typ, groupIdx)
  1012  			}
  1013  
  1014  			if groupMatch == 0 {
  1015  				groupMatch = it.group.ctrls().matchFull()
  1016  
  1017  				if slotIdx != 0 {
  1018  					// Starting in the middle of the group.
  1019  					// Ignore earlier groups.
  1020  					groupMatch = groupMatch.removeBelow(slotIdx)
  1021  				}
  1022  
  1023  				// Skip over groups that are composed of only empty or
  1024  				// deleted slots.
  1025  				if groupMatch == 0 {
  1026  					// Jump past remaining slots in this
  1027  					// group.
  1028  					it.entryIdx += abi.SwissMapGroupSlots - uint64(slotIdx)
  1029  					continue
  1030  				}
  1031  
  1032  				i := groupMatch.first()
  1033  				it.entryIdx += uint64(i - slotIdx)
  1034  				if it.entryIdx > entryMask {
  1035  					// Past the end of this table's iteration.
  1036  					continue
  1037  				}
  1038  				entryIdx += uint64(i - slotIdx)
  1039  				slotIdx = i
  1040  			}
  1041  
  1042  			key := it.group.key(it.typ, slotIdx)
  1043  			if it.typ.IndirectKey() {
  1044  				key = *((*unsafe.Pointer)(key))
  1045  			}
  1046  
  1047  			// If the table has changed since the last
  1048  			// call, then it has grown or split. In this
  1049  			// case, further mutations (changes to
  1050  			// key->elem or deletions) will not be visible
  1051  			// in our snapshot table. Instead we must
  1052  			// consult the new table by doing a full
  1053  			// lookup.
  1054  			//
  1055  			// We still use our old table to decide which
  1056  			// keys to lookup in order to avoid returning
  1057  			// the same key twice.
  1058  			grown := it.tab.index == -1
  1059  			var elem unsafe.Pointer
  1060  			if grown {
  1061  				newKey, newElem, ok := it.grownKeyElem(key, slotIdx)
  1062  				if !ok {
  1063  					// This entry doesn't exist anymore.
  1064  					// Continue to the next one.
  1065  					groupMatch = groupMatch.removeFirst()
  1066  					if groupMatch == 0 {
  1067  						// No more entries in this
  1068  						// group. Continue to next
  1069  						// group.
  1070  						it.entryIdx += abi.SwissMapGroupSlots - uint64(slotIdx)
  1071  						continue
  1072  					}
  1073  
  1074  					// Next full slot.
  1075  					i := groupMatch.first()
  1076  					it.entryIdx += uint64(i - slotIdx)
  1077  					continue
  1078  				} else {
  1079  					key = newKey
  1080  					elem = newElem
  1081  				}
  1082  			} else {
  1083  				elem = it.group.elem(it.typ, slotIdx)
  1084  				if it.typ.IndirectElem() {
  1085  					elem = *((*unsafe.Pointer)(elem))
  1086  				}
  1087  			}
  1088  
  1089  			// Jump ahead to the next full slot or next group.
  1090  			groupMatch = groupMatch.removeFirst()
  1091  			if groupMatch == 0 {
  1092  				// No more entries in
  1093  				// this group. Continue
  1094  				// to next group.
  1095  				it.entryIdx += abi.SwissMapGroupSlots - uint64(slotIdx)
  1096  			} else {
  1097  				// Next full slot.
  1098  				i := groupMatch.first()
  1099  				it.entryIdx += uint64(i - slotIdx)
  1100  			}
  1101  
  1102  			it.key = key
  1103  			it.elem = elem
  1104  			return
  1105  		}
  1106  
  1107  		// Continue to next table.
  1108  	}
  1109  
  1110  	it.key = nil
  1111  	it.elem = nil
  1112  	return
  1113  }
  1114  
  1115  // Replaces the table with one larger table or two split tables to fit more
  1116  // entries. Since the table is replaced, t is now stale and should not be
  1117  // modified.
  1118  func (t *table) rehash(typ *abi.SwissMapType, m *Map) {
  1119  	// TODO(prattmic): SwissTables typically perform a "rehash in place"
  1120  	// operation which recovers capacity consumed by tombstones without growing
  1121  	// the table by reordering slots as necessary to maintain the probe
  1122  	// invariant while eliminating all tombstones.
  1123  	//
  1124  	// However, it is unclear how to make rehash in place work with
  1125  	// iteration. Since iteration simply walks through all slots in order
  1126  	// (with random start offset), reordering the slots would break
  1127  	// iteration.
  1128  	//
  1129  	// As an alternative, we could do a "resize" to new groups allocation
  1130  	// of the same size. This would eliminate the tombstones, but using a
  1131  	// new allocation, so the existing grow support in iteration would
  1132  	// continue to work.
  1133  
  1134  	newCapacity := 2 * t.capacity
  1135  	if newCapacity <= maxTableCapacity {
  1136  		t.grow(typ, m, newCapacity)
  1137  		return
  1138  	}
  1139  
  1140  	t.split(typ, m)
  1141  }
  1142  
  1143  // Bitmask for the last selection bit at this depth.
  1144  func localDepthMask(localDepth uint8) uintptr {
  1145  	if goarch.PtrSize == 4 {
  1146  		return uintptr(1) << (32 - localDepth)
  1147  	}
  1148  	return uintptr(1) << (64 - localDepth)
  1149  }
  1150  
  1151  // split the table into two, installing the new tables in the map directory.
  1152  func (t *table) split(typ *abi.SwissMapType, m *Map) {
  1153  	localDepth := t.localDepth
  1154  	localDepth++
  1155  
  1156  	// TODO: is this the best capacity?
  1157  	left := newTable(typ, maxTableCapacity, -1, localDepth)
  1158  	right := newTable(typ, maxTableCapacity, -1, localDepth)
  1159  
  1160  	// Split in half at the localDepth bit from the top.
  1161  	mask := localDepthMask(localDepth)
  1162  
  1163  	for i := uint64(0); i <= t.groups.lengthMask; i++ {
  1164  		g := t.groups.group(typ, i)
  1165  		for j := uintptr(0); j < abi.SwissMapGroupSlots; j++ {
  1166  			if (g.ctrls().get(j) & ctrlEmpty) == ctrlEmpty {
  1167  				// Empty or deleted
  1168  				continue
  1169  			}
  1170  
  1171  			key := g.key(typ, j)
  1172  			if typ.IndirectKey() {
  1173  				key = *((*unsafe.Pointer)(key))
  1174  			}
  1175  
  1176  			elem := g.elem(typ, j)
  1177  			if typ.IndirectElem() {
  1178  				elem = *((*unsafe.Pointer)(elem))
  1179  			}
  1180  
  1181  			hash := typ.Hasher(key, m.seed)
  1182  			var newTable *table
  1183  			if hash&mask == 0 {
  1184  				newTable = left
  1185  			} else {
  1186  				newTable = right
  1187  			}
  1188  			newTable.uncheckedPutSlot(typ, hash, key, elem)
  1189  		}
  1190  	}
  1191  
  1192  	m.installTableSplit(t, left, right)
  1193  	t.index = -1
  1194  }
  1195  
  1196  // grow the capacity of the table by allocating a new table with a bigger array
  1197  // and uncheckedPutting each element of the table into the new table (we know
  1198  // that no insertion here will Put an already-present value), and discard the
  1199  // old table.
  1200  func (t *table) grow(typ *abi.SwissMapType, m *Map, newCapacity uint16) {
  1201  	newTable := newTable(typ, uint64(newCapacity), t.index, t.localDepth)
  1202  
  1203  	if t.capacity > 0 {
  1204  		for i := uint64(0); i <= t.groups.lengthMask; i++ {
  1205  			g := t.groups.group(typ, i)
  1206  			for j := uintptr(0); j < abi.SwissMapGroupSlots; j++ {
  1207  				if (g.ctrls().get(j) & ctrlEmpty) == ctrlEmpty {
  1208  					// Empty or deleted
  1209  					continue
  1210  				}
  1211  
  1212  				key := g.key(typ, j)
  1213  				if typ.IndirectKey() {
  1214  					key = *((*unsafe.Pointer)(key))
  1215  				}
  1216  
  1217  				elem := g.elem(typ, j)
  1218  				if typ.IndirectElem() {
  1219  					elem = *((*unsafe.Pointer)(elem))
  1220  				}
  1221  
  1222  				hash := typ.Hasher(key, m.seed)
  1223  
  1224  				newTable.uncheckedPutSlot(typ, hash, key, elem)
  1225  			}
  1226  		}
  1227  	}
  1228  
  1229  	newTable.checkInvariants(typ, m)
  1230  	m.replaceTable(newTable)
  1231  	t.index = -1
  1232  }
  1233  
  1234  // probeSeq maintains the state for a probe sequence that iterates through the
  1235  // groups in a table. The sequence is a triangular progression of the form
  1236  //
  1237  //	p(i) := (i^2 + i)/2 + hash (mod mask+1)
  1238  //
  1239  // The sequence effectively outputs the indexes of *groups*. The group
  1240  // machinery allows us to check an entire group with minimal branching.
  1241  //
  1242  // It turns out that this probe sequence visits every group exactly once if
  1243  // the number of groups is a power of two, since (i^2+i)/2 is a bijection in
  1244  // Z/(2^m). See https://en.wikipedia.org/wiki/Quadratic_probing
  1245  type probeSeq struct {
  1246  	mask   uint64
  1247  	offset uint64
  1248  	index  uint64
  1249  }
  1250  
  1251  func makeProbeSeq(hash uintptr, mask uint64) probeSeq {
  1252  	return probeSeq{
  1253  		mask:   mask,
  1254  		offset: uint64(hash) & mask,
  1255  		index:  0,
  1256  	}
  1257  }
  1258  
  1259  func (s probeSeq) next() probeSeq {
  1260  	s.index++
  1261  	s.offset = (s.offset + s.index) & s.mask
  1262  	return s
  1263  }
  1264  
  1265  func (t *table) clone(typ *abi.SwissMapType) *table {
  1266  	// Shallow copy the table structure.
  1267  	t2 := new(table)
  1268  	*t2 = *t
  1269  	t = t2
  1270  
  1271  	// We need to just deep copy the groups.data field.
  1272  	oldGroups := t.groups
  1273  	newGroups := newGroups(typ, oldGroups.lengthMask+1)
  1274  	for i := uint64(0); i <= oldGroups.lengthMask; i++ {
  1275  		oldGroup := oldGroups.group(typ, i)
  1276  		newGroup := newGroups.group(typ, i)
  1277  		cloneGroup(typ, newGroup, oldGroup)
  1278  	}
  1279  	t.groups = newGroups
  1280  
  1281  	return t
  1282  }
  1283
View as plain text