increase telemetry granularity

breandan · Apr 16, 2024 · 2ab779c · 2ab779c
1 parent 3570a4e
commit 2ab779c
Show file tree

Hide file tree

Showing 3 changed files with 32 additions and 14 deletions.
diff --git a/src/commonMain/kotlin/ai/hypergraph/kaliningraph/parsing/BarHillel.kt b/src/commonMain/kotlin/ai/hypergraph/kaliningraph/parsing/BarHillel.kt
@@ -62,6 +62,8 @@ private infix fun CFG.intersectLevFSAP(fsa: FSA): CFG {
   ct.filter { fsa.obeys(it.π1, it.π2, it.π3, parikhMap) }
     .forEach { ct2[it.π1.π1][it.π3][it.π2.π1] = true }
 
+  val states = fsa.stateLst
+  val allsym = ntLst
   val binaryProds =
     prods.map {
 //      if (i % 100 == 0) println("Finished ${i}/${nonterminalProductions.size} productions")
@@ -74,8 +76,8 @@ private infix fun CFG.intersectLevFSAP(fsa: FSA): CFG {
         .filter { it.checkCompatibility(trip, ct2) }
 //        .filter { it.obeysLevenshteinParikhBounds(A to B to C, fsa, parikhMap) }
         .map { (a, b, c) ->
-          val (p, q, r)  = fsa.stateLst[a] to fsa.stateLst[b] to fsa.stateLst[c]
-          "[$p~${ntLst[A]}~$r]".also { nts.add(it) } to listOf("[$p~${ntLst[B]}~$q]", "[$q~${ntLst[C]}~$r]")
+          val (p, q, r)  = states[a] to states[b] to states[c]
+          "[$p~${allsym[A]}~$r]".also { nts.add(it) } to listOf("[$p~${allsym[B]}~$q]", "[$q~${allsym[C]}~$r]")
         }.toList()
     }.flatten().filterRHSInNTS()
 

diff --git a/src/commonMain/kotlin/ai/hypergraph/kaliningraph/parsing/CFG.kt b/src/commonMain/kotlin/ai/hypergraph/kaliningraph/parsing/CFG.kt
@@ -7,6 +7,7 @@ import ai.hypergraph.kaliningraph.tokenizeByWhitespace
 import ai.hypergraph.kaliningraph.types.*
 import kotlin.jvm.JvmName
 import kotlin.time.*
+import kotlin.time.Duration.Companion.seconds
 
 typealias Σᐩ = String
 typealias Production = Π2<Σᐩ, List<Σᐩ>>

diff --git a/src/jvmMain/kotlin/ai/hypergraph/kaliningraph/parsing/JVMBarHillel.kt b/src/jvmMain/kotlin/ai/hypergraph/kaliningraph/parsing/JVMBarHillel.kt
@@ -167,19 +167,24 @@ private fun CFG.jvmIntersectLevFSAP(fsa: FSA, parikhMap: ParikhMap): CFG {
   // we have the production [p,A,r] → [p,B,q] [q,C,r] in P′.
   val prods = nonterminalProductions
     .map { (a, b) -> ntMap[a]!! to b.map { ntMap[it]!! } }.toSet()
-//  val lengthBoundsCache = lengthBounds.let { lb -> nonterminals.map { lb[it] ?: 0..0 } }
+  val lengthBoundsCache = lengthBounds.let { lb -> nonterminals.map { lb[it] ?: 0..0 } }
   val validTriples = fsa.validTriples.map { arrayOf(it.π1.π1, it.π2.π1, it.π3.π1) }
 
+  val ctClock = TimeSource.Monotonic.markNow()
   val ct = (fsa.validPairs * nonterminals.indices.toSet()).toList()
   val ct2 = Array(fsa.states.size) { Array(nonterminals.size) { Array(fsa.states.size) { false } } }
   ct.parallelStream()
-    .filter { fsa.obeys(it.π1, it.π2, it.π3, parikhMap) }
-    .toList().also {
+    .filter {
+      lengthBoundsCache[it.π3].overlaps(fsa.SPLP(it.π1, it.π2)) &&
+        fsa.obeys(it.π1, it.π2, it.π3, parikhMap)
+    }.toList().also {
       val fraction = it.size.toDouble() / (fsa.states.size * nonterminals.size * fsa.states.size)
       println("Fraction of valid triples: $fraction")
     }.forEach { ct2[it.π1.π1][it.π3][it.π2.π1] = true }
+  println("Precomputed LP constraints in ${ctClock.elapsedNow()}")
 
-  val elimCounter = AtomicInteger(0)
+  val states = fsa.stateLst
+  val allsym = ntLst
   val counter = AtomicInteger(0)
   val lpClock = TimeSource.Monotonic.markNow()
   val binaryProds =
@@ -190,17 +195,18 @@ private fun CFG.jvmIntersectLevFSAP(fsa: FSA, parikhMap: ParikhMap): CFG {
       validTriples.stream()
         // CFG ∩ FSA - in general we are not allowed to do this, but it works
         // because we assume a Levenshtein FSA, which is monotone and acyclic.
-//        .filter { it.isCompatibleWith(A to B to C, fsa, lengthBoundsCache).also { if (!it) elimCounter.incrementAndGet() } }
+//        .filter { it.isCompatibleWith(A to B to C, fsa, lengthBoundsCache) }
 //        .filter { it.checkCT(trip, ct1).also { if (!it) elimCounter.incrementAndGet() } }
-//        .filter { it.obeysLevenshteinParikhBounds(A to B to C, fsa, parikhMap).also { if (!it) elimCounter.incrementAndGet() } }
-        .filter { it.checkCompatibility(trip, ct2).also { if (!it) elimCounter.incrementAndGet() } }
+//        .filter { it.obeysLevenshteinParikhBounds(A to B to C, fsa, parikhMap) }
+        .filter { it.checkCompatibility(trip, ct2) }
         .map { (a, b, c) ->
           if (MAX_PRODS < counter.incrementAndGet()) throw Exception("∩-grammar has too many productions! (>$MAX_PRODS)")
-          val (p, q, r)  = fsa.stateLst[a] to fsa.stateLst[b] to fsa.stateLst[c]
-          "[$p~${ntLst[A]}~$r]".also { nts.add(it) } to listOf("[$p~${ntLst[B]}~$q]", "[$q~${ntLst[C]}~$r]")
+          val (p, q, r)  = states[a] to states[b] to states[c]
+          "[$p~${allsym[A]}~$r]".also { nts.add(it) } to listOf("[$p~${allsym[B]}~$q]", "[$q~${allsym[C]}~$r]")
         }
     }.toList()
 
+  val elimCounter = (validTriples.size * prods.size) - binaryProds.size
   println("Levenshtein-Parikh constraints eliminated $elimCounter productions in ${lpClock.elapsedNow()}")
 
   fun Σᐩ.isSyntheticNT() =
@@ -213,6 +219,7 @@ private fun CFG.jvmIntersectLevFSAP(fsa: FSA, parikhMap: ParikhMap): CFG {
   return Stream.concat(binaryProds.stream(), (initFinal + transits + unitProds).stream()).parallel()
     .filter { (_, rhs) -> rhs.all { !it.isSyntheticNT() || it in nts } }
     .collect(Collectors.toSet())
+    .also { println("Eliminated ${totalProds - it.size} extra productions before normalization") }
     .jvmPostProcess(clock)
 }
 
@@ -249,15 +256,15 @@ fun CFG.jvmDropVestigialProductions(clock: TimeSource.Monotonic.ValueTimeMark):
   val nts: Set<Σᐩ> = asSequence().asStream().parallel().map { it.first }.collect(Collectors.toSet())
   val rw = asSequence().asStream().parallel()
     .filter { prod ->
-      if (counter.incrementAndGet() % 10 == 0 && BH_TIMEOUT < clock.elapsedNow()) throw Exception("Timeout!")
+     if (counter.incrementAndGet() % 10 == 0 && BH_TIMEOUT < clock.elapsedNow()) throw Exception("Timeout! ${clock.elapsedNow()}")
       // Only keep productions whose RHS symbols are not synthetic or are in the set of NTs
       prod.RHS.all { !(it.first() == '[' && 1 < it.length) || it in nts }
     }
     .collect(Collectors.toSet())
-//    .also { println("Removed ${size - it.size} invalid productions in ${clock.elapsedNow() - start}") }
+    .also { println("Removed ${size - it.size} invalid productions in ${clock.elapsedNow() - start}") }
     .freeze().jvmRemoveUselessSymbols()
 
-//  println("Removed ${size - rw.size} vestigial productions, resulting in ${rw.size} productions.")
+  println("Removed ${size - rw.size} vestigial productions, resulting in ${rw.size} productions.")
 
   return if (rw.size == size) rw else rw.jvmDropVestigialProductions(clock)
 }
@@ -293,6 +300,10 @@ private fun CFG.jvmReachSym(from: Σᐩ = START_SYMBOL): Set<Σᐩ> {
       this@jvmReachSym.asSequence().asStream().parallel()
         .forEach { (l, r) -> getOrPut(l) { ConcurrentSkipListSet() }.addAll(r) }
     }
+//    [email protected]().asStream().parallel()
+//      .flatMap { (l, r) -> r.stream().map { l to it } }
+//      // List of second elements grouped by first element
+//      .collect(Collectors.groupingByConcurrent({ it.first }, Collectors.mapping({ it.second }, Collectors.toSet())))
 
   do {
     val t = nextReachable.first()
@@ -320,6 +331,10 @@ private fun CFG.jvmGenSym(
       this@jvmGenSym.asSequence().asStream().parallel()
         .forEach { (l, r) -> r.forEach { getOrPut(it) { ConcurrentSkipListSet() }.add(l) } }
     }
+//    [email protected]().asStream().parallel()
+//      .flatMap { (l, r) -> r.asSequence().asStream().map { it to l } }
+//      // List of second elements grouped by first element
+//      .collect(Collectors.groupingByConcurrent({ it.first }, Collectors.mapping({ it.second }, Collectors.toList())))
 
   do {
     val t = nextGenerating.first()