Skip to content

perf(hermitcrab): run the required-slot look-ahead on the shared trailing-slot template walk #577

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
johnml1135 wants to merge 8 commits into
base: master
Choose a base branch
from
Draft

perf(hermitcrab): run the required-slot look-ahead on the shared trailing-slot template walk #577

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
0a7ec6f
perf(hermitcrab): share trailing affix-template slot batches
johnml1135 Oct 22, 2026
c07febe
Address review: drop the unreachable limit check, explain the fallbac…
johnml1135 Oct 22, 2026
1c51b2c
Shorten the shared-slot fallback comment to the hygiene budget
johnml1135 Oct 23, 2026
17a6894
Compare shared-walk outputs with AnalysisMergeKey
johnml1135 Oct 23, 2026
4724e4d
Prune affix template partials whose required slot cannot find its mat…
johnml1135 Oct 28, 2026
fcb65d7
Run the required-slot look-ahead on the shared trailing-slot template…
johnml1135 Oct 28, 2026
645a795
The shared look-ahead walk no longer re-scans a word after skipping a…
johnml1135 Nov 12, 2026
c8df81a
Shorten the look-ahead skip comment to meet the comment-hygiene budget.
johnml1135 Nov 12, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
146 changes: 126 additions & 20 deletions src/SIL.Machine.Morphology.HermitCrab/AnalysisAffixTemplateRule.cs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,7 @@ internal class AnalysisAffixTemplateRule : IRule<Word, ShapeNode>
private readonly Morpher _morpher;
private readonly AffixTemplate _template;
private readonly List<IRule<Word, ShapeNode>> _rules;
private readonly RequiredSlotMaterial _lookahead;

public AnalysisAffixTemplateRule(Morpher morpher, AffixTemplate template)
{
Expand All @@ -26,17 +27,47 @@ public AnalysisAffixTemplateRule(Morpher morpher, AffixTemplate template)
FreezableEqualityComparer<Word>.Default
))
);
_lookahead = new RequiredSlotMaterial(template);
}

public IEnumerable<Word> Apply(Word input)
{
if (!_morpher.RuleSelector(_template))
if (!CanApply(input))
return Enumerable.Empty<Word>();

int[] fired = null;
if (RequiredSlotCannotUnapply(input, _rules.Count - 1, ref fired))
{
if (_morpher.TraceManager.IsTracing)
_morpher.TraceManager.EndUnapplyTemplate(_template, input, false, FailureReason.None);
return Enumerable.Empty<Word>();
}

Word inWord = input.Clone();
inWord.Freeze();

var output = new HashSet<Word>(FreezableEqualityComparer<Word>.Default);
if (_morpher.MaxDegreeOfParallelism == 1)
ApplySlots(inWord, _rules.Count - 1, output, fired);
else
ParallelApplySlots(inWord, output, fired);

return output;
}

internal AffixTemplate Template => _template;

internal IReadOnlyList<IRule<Word, ShapeNode>> SlotRules => _rules;

internal bool CanApply(Word input)
{
if (!_morpher.RuleSelector(_template))
return false;

// A check only: adding the template's required features here would make SyntacticFeatureStruct
// depend on whether a template was un-applied, which Word.ValueEquals assumes cannot happen.
if (!input.SyntacticFeatureStruct.IsUnifiable(_template.RequiredSyntacticFeatureStruct))
return Enumerable.Empty<Word>();
return false;

if (_morpher.TraceManager.IsTracing)
_morpher.TraceManager.BeginUnapplyTemplate(_template, input);
Expand All @@ -58,27 +89,84 @@ public IEnumerable<Word> Apply(Word input)
FailureReason.NonPartialRuleProhibitedAfterFinalTemplate
);
}
return Enumerable.Empty<Word>();
return false;
}

Word inWord = input.Clone();
inWord.Freeze();
return true;
}

var output = new HashSet<Word>(FreezableEqualityComparer<Word>.Default);
if (_morpher.MaxDegreeOfParallelism == 1)
ApplySlots(inWord, _rules.Count - 1, output);
else
ParallelApplySlots(inWord, output);
// Testing runs from the next slot to unapply and stops after an impure slot, which makes the test
// for any slot behind it unsound. Shadow mode prunes nothing and carries the fired slots in fired.
internal bool RequiredSlotCannotUnapply(Word word, int lastSlot, ref int[] fired)
{
RequiredSlotLookaheadMode mode = _morpher.RequiredSlotLookahead;
if (mode == RequiredSlotLookaheadMode.Off)
return false;

return output;
long unifications = 0;
bool cannot = false;
for (int j = lastSlot; j >= 0; j--)
{
if (
!_template.Slots[j].Optional
&& !IsFired(fired, j)
&& _lookahead.CannotUnapply(j, word.Shape, ref unifications)
)
{
if (mode == RequiredSlotLookaheadMode.On)
{
cannot = true;
break;
}
fired = fired == null ? new[] { j } : fired.Append(j).ToArray();
_morpher.RequiredSlotShadow?.Invoke("fired", _template, j);
}
if (!_lookahead.IsPure(j))
break;
}
if (unifications != 0)
_morpher.RequiredSlotCheckObserver?.Invoke(unifications);
return cannot;
}

internal bool IsSlotPure(int slot) => _lookahead.IsPure(slot);

internal static bool IsFired(int[] fired, int slot)
{
return fired != null && Array.IndexOf(fired, slot) >= 0;
}

private void ApplySlots(Word inWord, int index, HashSet<Word> output)
internal void ReportShadow(string kind, int[] fired, int slot)
{
if (fired == null || _morpher.RequiredSlotShadow == null)
return;
if (slot < 0)
{
foreach (int j in fired)
_morpher.RequiredSlotShadow(kind, _template, j);
}
else if (IsFired(fired, slot))
{
_morpher.RequiredSlotShadow(kind, _template, slot);
}
}

private void ApplySlots(Word inWord, int index, HashSet<Word> output, int[] fired)
{
for (int i = index; i >= 0; i--)
{
foreach (Word outWord in _rules[i].Apply(inWord))
ApplySlots(outWord, i - 1, output);
{
ReportShadow("violated", fired, i);
int[] outFired = fired;
if (RequiredSlotCannotUnapply(outWord, i - 1, ref outFired))
{
if (_morpher.TraceManager.IsTracing)
_morpher.TraceManager.EndUnapplyTemplate(_template, outWord, false, FailureReason.None);
continue;
}
ApplySlots(outWord, i - 1, output, outFired);
}

if (!_template.Slots[i].Optional)
{
Expand All @@ -90,16 +178,17 @@ private void ApplySlots(Word inWord, int index, HashSet<Word> output)

if (_morpher.TraceManager.IsTracing)
_morpher.TraceManager.EndUnapplyTemplate(_template, inWord, true, FailureReason.None);
ReportShadow("completed-under-fired", fired, -1);
output.Add(inWord);
}

private void ParallelApplySlots(Word inWord, HashSet<Word> output)
private void ParallelApplySlots(Word inWord, HashSet<Word> output, int[] fired)
{
ParallelOptions parallelOptions = _morpher.CreateParallelOptions();
var outStack = new ConcurrentStack<Word>();
var from = new ConcurrentStack<Tuple<Word, int>>();
from.Push(Tuple.Create(inWord, _rules.Count - 1));
var to = new ConcurrentStack<Tuple<Word, int>>();
var from = new ConcurrentStack<Tuple<Word, int, int[]>>();
from.Push(Tuple.Create(inWord, _rules.Count - 1, fired));
var to = new ConcurrentStack<Tuple<Word, int, int[]>>();
while (!from.IsEmpty)
{
to.Clear();
Expand All @@ -111,9 +200,25 @@ private void ParallelApplySlots(Word inWord, HashSet<Word> output)
bool add = true;
for (int i = work.Item2; i >= 0; i--)
{
Tuple<Word, int>[] workItems = _rules[i]
int next = i - 1;
Tuple<Word, int, int[]>[] workItems = _rules[i]
.Apply(work.Item1)
.Select(res => Tuple.Create(res, i - 1))
.Select(res =>
{
ReportShadow("violated", work.Item3, i);
int[] resFired = work.Item3;
if (!RequiredSlotCannotUnapply(res, next, ref resFired))
return Tuple.Create(res, next, resFired);
if (_morpher.TraceManager.IsTracing)
_morpher.TraceManager.EndUnapplyTemplate(
_template,
res,
false,
FailureReason.None
);
return null;
})
.Where(item => item != null)
.ToArray();
if (workItems.Length > 0)
to.PushRange(workItems);
Expand Down Expand Up @@ -141,11 +246,12 @@ private void ParallelApplySlots(Word inWord, HashSet<Word> output)
true,
FailureReason.None
);
ReportShadow("completed-under-fired", work.Item3, -1);
outStack.Push(work.Item1);
}
}
);
ConcurrentStack<Tuple<Word, int>> temp = from;
ConcurrentStack<Tuple<Word, int, int[]>> temp = from;
from = to;
to = temp;
}
Expand Down
Loading
Loading