Changeset 1231 for cpp/frams/genetics

Timestamp:

05/02/23 01:36:15 (19 months ago)

Author:

Maciej Komosinski

Message:

• Thanks to r1230, it is possible to detect (and repair=remove) junk trailing genes that are left after successful parsing (after last '>')
• The validate() function may attempt to repair a genotype where earlier it would give up
• Stricter parsing of the '#' gene

Location:

cpp/frams/genetics/f4

Files:

• f4_conv.cpp (modified) (7 diffs)
• f4_general.cpp (modified) (8 diffs)
• f4_general.h (modified) (4 diffs)
• f4_oper.cpp (modified) (8 diffs)
• f4_oper.h (modified) (1 diff)

cpp/frams/genetics/f4/f4_conv.cpp

@@ -27 +27 @@
 SString GenoConv_f40::convert(SString &in, MultiMap *map, bool using_checkpoints)
 {
-        int res;
         f4_Model *model = new f4_Model();
-        res = model->buildFromF4(in, using_checkpoints);
-        if (GENOPER_OK != res)
+        int res = model->buildFromF4(in, using_checkpoints);
+        if (res != GENOPER_OK)
         {
                 delete model;
@@ -55 +54 @@
 SString GenoConv_F41_TestOnly::convert(SString &in, MultiMap *map, bool using_checkpoints)
 {
-        int res;
         f4_Model *model = new f4_Model();
-        res = model->buildFromF4(in, using_checkpoints);
-        if (GENOPER_OK != res)
+        int res = model->buildFromF4(in, using_checkpoints);
+        if (res != GENOPER_OK)
         {
                 delete model;
@@ -82 +80 @@
 int f4_Model::buildFromF4(SString &geno, bool using_checkpoints)
 {
-        int i;
-
         error = GENOPER_OK;
         errorpos = -1;
 
+        // transform geno from string to nodes
+        f4_Node f4rootnode;
+        int res = f4_process(geno.c_str(), &f4rootnode);
+        if (res || (f4rootnode.childCount() != 1)) //consider any error fatal, preventing building a model
+        {
+                error = GENOPER_OPFAIL;
+                errorpos = res;
+                return error;
+        }
+
         // build cells, and simulate
         if (cells) delete cells;
-        cells = new f4_Cells(geno, 0);
-        if (GENOPER_OK != cells->getErrorCode())
+        cells = new f4_Cells(f4rootnode.child, false);
+        if (cells->getErrorCode() != GENOPER_OK)
         {
                 error = cells->getErrorCode();
@@ -99 +105 @@
 
         cells->simulate();
-        if (GENOPER_OK != cells->getErrorCode())
+        if (cells->getErrorCode() != GENOPER_OK)
         {
                 error = cells->getErrorCode();
@@ -107 +113 @@
 
         // reset recursive traverse flags
-        for (i = 0; i < cells->cell_count; i++)
+        for (int i = 0; i < cells->cell_count; i++)
                 cells->C[i]->recProcessedFlag = 0;
 
@@ -113 +119 @@
 
         // process every cell
-        int res;
-        for (i = 0; i < cells->cell_count; i++)
-        {
-                res = buildModelRec(cells->C[i]);
+        for (int i = 0; i < cells->cell_count; i++)
+        {
+                int res = buildModelRec(cells->C[i]);
                 if (res)
                 {
-                        logMessage("f4_Model", "buildFromF4", LOG_ERROR, "Error in building a Model");
+                        logPrintf("f4_Model", "buildFromF4", LOG_ERROR, "Error %d when building a Model", res);
                         error = res;
                         break;
@@ -125 +130 @@
         }
 
-        res = close();
-        if (0 == res) // invalid
+        int res_close = close();
+        if (res_close == 0) // invalid
+        {
+                logPrintf("f4_Model", "buildFromF4", LOG_ERROR, "Error %d when closing a Model", res_close);
                 error = -10;
+        }
 
         return error;

cpp/frams/genetics/f4/f4_general.cpp

@@ -640 +640 @@
 
 
-f4_Cells::f4_Cells(f4_Node *genome, int nrepair)
-{
-        // create ancestor cell
+f4_Cells::f4_Cells(f4_Node *genome, bool nrepair)
+{
         repair = nrepair;
         errorcode = GENOPER_OK;
@@ -650 +649 @@
         repair_insert = NULL;
         tmpcel = NULL;
-        f4rootnode = NULL;
+
+        // create ancestor cell
         C[0] = new f4_Cell(this, 0, genome, genome, NULL, 0, GeneProps::standard_values);
         cell_count = 1;
@@ -656 +656 @@
 
 
-f4_Cells::f4_Cells(SString & genome, int nrepair)
-{
-        repair = nrepair;
-        errorcode = GENOPER_OK;
-        errorpos = -1;
-        repair_remove = NULL;
-        repair_parent = NULL;
-        repair_insert = NULL;
-        tmpcel = NULL;
-        f4rootnode = NULL;
-
-        // transform geno from string to nodes
-        f4rootnode = new f4_Node();
-        int _ = 0;
-        int res = f4_processRecur(genome.c_str(), _, f4rootnode);
-        if (res || (f4rootnode->childCount() != 1))
-        {
-                errorcode = GENOPER_OPFAIL;
-                errorpos = -1;
-        }
-
-        // create ancestor cell
-        C[0] = new f4_Cell(this, 0, f4rootnode->child, f4rootnode->child, NULL, 0, GeneProps::standard_values);
-        cell_count = 1;
-}
 
 f4_Cells::~f4_Cells()
 {
         // release cells
-        int i;
         if (cell_count)
         {
-                for (i = cell_count - 1; i >= 0; i--)
+                for (int i = cell_count - 1; i >= 0; i--)
                         delete C[i];
                 cell_count = 0;
         }
-        if (f4rootnode)
-                delete f4rootnode;
 }
 
@@ -1130 +1102 @@
 int f4_Node::childCount()
 {
-        if (child != NULL)
-        {
-                if (child2 != NULL) return 2;
-                else return 1;
-        }
-        else
-        {
-                if (child2 != NULL) return 1;
-                else return 0;
-        }
+        return int(child != NULL) + int(child2 != NULL); //0, 1 or 2
 }
 
@@ -1256 +1219 @@
         len = out.length();
         if (len > 1)
-                if (out[len - 1] == '>') { (out.directWrite())[len - 1] = 0; out.endWrite(); }; //Macko 2023-04 TODO "can be omitted", but should we remove it as a rule even in generated genotypes? see if I can somehow detect junk characters after top-level '>' ends properly: /*4*/<X>N:N>whatever
+                if (out[len - 1] == '>') { (out.directWrite())[len - 1] = 0; out.endWrite(); }; //Macko 2023-04 "can be omitted", but it is removed as a rule even in generated genotypes :)
         // copy back to string
         // if new is longer, reallocate buf
@@ -1305 +1268 @@
         while (pos_inout < (int)strlen(genot))
         {
-                //#define PRINT_PARSING_LOCATION
+//#define PRINT_PARSING_LOCATION
 #ifdef PRINT_PARSING_LOCATION
                 printf("%s\n", genot);
@@ -1342 +1305 @@
                 case '#':
                 {
-                        // repetition marker, 1 by default
-                        ExtValue val;
-                        const char* end = val.parseNumber(genot + pos_inout + 1, ExtPType::TInt);
-                        //TODO end==NULL? -> error!
-                        int reps = (end == NULL) ? 1 : val.getInt();
+                        // repetition marker
+                        ExtValue reps;
+                        const char* end = reps.parseNumber(genot + pos_inout + 1, ExtPType::TInt);
+                        if (end == NULL)
+                                return pos_inout + 1; //error
                         f4_Node *node = new f4_Node("#", par, pos_inout);
-                        node->reps = reps;
+                        node->reps = reps.getInt();
                         // skip number
                         pos_inout += end - (genot + pos_inout);
@@ -1455 +1418 @@
 }
 
+int f4_process(const char *genot, f4_Node *root)
+{
+        int pos = 0;
+        int res = f4_processRecur(genot, pos, root);
+        if (res > 0)
+                return res; //error
+        else if (genot[pos] == 0) //parsed until the end - OK!
+                return 0;
+        else return pos + 1; //junk, unparsed genes after successful parsing, for example /*4*/<X>N:N>whatever or /*4*/<X>X>>>
+}
+
 const char* parseConnection(const char *fragm, int& relfrom, double &weight)
 {

cpp/frams/genetics/f4/f4_general.h

@@ -242 +242 @@
          * Constructor taking genotype in a form of a tree.
          * @param genome genotype tree
-         * @param nrepair 0 if nothing to repair
-         */
-        f4_Cells(f4_Node *genome, int nrepair);
-
-        /**
-         * Constructor taking genotype in a form of a string.
-         * @param genome genotype string
-         * @param nrepair 0 if nothing to repair
-         */
-        f4_Cells(SString &genome, int nrepair);
+         * @param nrepair false if nothing to repair
+         */
+        f4_Cells(f4_Node *genome, bool nrepair);
 
         /**
@@ -336 +329 @@
 private:
         // for error reporting / genotype fixing
-        int repair;
+        bool repair;
         int errorcode;
         int errorpos;
@@ -343 +336 @@
         f4_Node *repair_insert;
         void toF1GenoRec(int curc, SString &out);
-        f4_Cell *tmpcel;                // needed by toF1Geno
-        f4_Node *f4rootnode;          // used by constructor
+        f4_Cell *tmpcel;  // needed by toF1Geno
 };
 
@@ -477 +469 @@
  * semantic aspect is neuron class name extraction, where the GenoOperators
  * class is used to parse the potential neuron class name.
- * @param genot the string holding all the genotype
- * @param pos0 the current position of processing in string
+ * This is an internal function; for regular cases, use f4_process().
+ * @param genot the string with the entire genotype
+ * @param pos_inout the current position of processing in string (advanced by the function)
  * @param parent current parent of the analysed branch of the genotype
  * @return 0 if processing was successful, otherwise returns the position of an error in the genotype
  */
 int f4_processRecur(const char *genot, int &pos_inout, f4_Node *parent);
+
+/**
+ * A wrapper for f4_processRecur(). Creates a tree of f4_Node objects corresponding to
+ * the provided genotype.
+ * @param genot the string with the entire genotype
+ * @param root root of the tree corresponding to the genotype
+ * @return 0 if processing was successful, otherwise returns the position of an error in the genotype
+ */
+int f4_process(const char *genot, f4_Node *root);
 
 /**

cpp/frams/genetics/f4/f4_oper.cpp

@@ -11 +11 @@
 // may help, but it would be better to improve the source code to make genetic operators neutral in terms of genotype length. Adding such a penalty
 // removes "work in progress" changes in genotypes thus promoting immediate, straightforward improvements while hindering slower, multifaceted progress.
-// TODO getting rid of redundancy (having valid genotypes with a lot of "junk code") in this representation looks like a good idea.
+// TODO getting rid of redundancy (valid genotypes with a lot of "junk code") in this representation looks like a good idea; many improvements to this end have already been done in April & May 2023.
 // 
-// Note: symbols after the last > are ignored, for example /*4*/<X>N:N>blablaN:N[2:-0.5]XXXwhatever but since they are not parsed into the f4_Node tree, they will be lost after any mutation.
 //
 // TODO the behavior of neuron input indexes during mutation seems badly implemented (see also TREAT_BAD_CONNECTIONS_AS_INVALID_GENO). Are they kept properly maintained when nodes are added and removed? This could be done well because during mutation we operate on the tree structure with cross-references between nodes (so they should not be affected by local changes in the tree), and then convert the tree back to string. Yet, the f4_Node.conn_from is an integer and these fields in nodes do not seem to be maintained on tree node adding/removal... change these integer offsets to references to node objects? But actually, do the offsets that constitute relative connection references concern the f4_Node tree structure (and all these sophisticated calculations of offsets during mutation are useful) or rather they concern the f4_Cells development? verify all situations in f4_Cell::oneStep(), case '['.
@@ -87 +86 @@
 }
 
-int Geno_f4::ValidateRec(f4_Node *geno, int retrycount) const
+int Geno_f4::ValidateRecur(f4_Node *geno, int retrycount) const
 {
         // ! the genotype is geno->child (not geno) !
         // build from it with repair on
 
-        f4_Cells cells(geno->child, 1);
+        f4_Cells cells(geno->child, true);
         cells.simulate();  //we should simulate?!
 
@@ -109 +108 @@
                 int res2 = GENOPER_OK;
                 if (retrycount > 0)
-                        res2 = ValidateRec(geno, retrycount - 1);
+                        res2 = ValidateRecur(geno, retrycount - 1);
 
                 if (res2 == GENOPER_OK) return GENOPER_REPAIR;
@@ -123 +122 @@
         // convert geno to a tree, then try to validate
         f4_Node root;
-        int _ = 0;
-        if (f4_processRecur(geno, _, &root) || root.childCount() != 1) return GENOPER_OK; // cannot repair
-
+        int res = f4_process(geno, &root);
+        if (res == 0 || root.childCount() != 1) return GENOPER_OK; // either parsing says the genotype is OK or the resulting tree will not be repairable (fatal flaw; root must have exactly one child) - do not attempt repair
+
+        // here we have a genotype with res>0 (for sure has some error) and root.childCount()==1 (still something was parsed into a tree)
         const int VALIDATE_TRIALS = 20;
-        if (ValidateRec(&root, VALIDATE_TRIALS) == GENOPER_REPAIR) // if repaired, make it back to string
+        res = ValidateRecur(&root, VALIDATE_TRIALS);
+        if (res != GENOPER_OPFAIL) // if repaired (GENOPER_REPAIR) or had no errors (GENOPER_OK, e.g. the genotype had some errors that were ignored during tree creation or had junk genes appended at the end, so the tree was OK but the genotype was not), 
         {
                 geno[0] = 0;
-                root.child->sprintAdj(geno);
+                root.child->sprintAdj(geno); //make it back to string
         }
         return GENOPER_OK;
@@ -139 +140 @@
 {
         f4_Node root;
-        int _ = 0;
-        int res = f4_processRecur(geno, _, &root);
+        int res = f4_process(geno, &root);
         if (res) return res;  // errorpos, >0
-        if (root.childCount() != 1) return 1; //earlier: GENOPER_OPFAIL
-        f4_Cells cells(root.child, 0);
+        if (root.childCount() != 1) return 1; // fatal flaw; root must have exactly one child
+        f4_Cells cells(root.child, false);
         cells.simulate();
         if (cells.getErrorCode() == GENOPER_OPFAIL || cells.getErrorCode() == GENOPER_REPAIR)
         {
                 if (cells.getErrorPos() >= 0) return 1 + cells.getErrorPos();
-                else return 1; //earlier: GENOPER_OPFAIL;
+                else return 1; //error, no known position
         }
         else return GENOPER_OK;
@@ -525 +525 @@
                 }
                 // try to validate it
-                res = ValidateRec(gcopy, 10);
+                res = ValidateRecur(gcopy, 10);
                 // accept if it is OK, or was repaired
                 if (GENOPER_OK == res)
@@ -552 +552 @@
 {
         f4_Node *root = new f4_Node;
-        int _ = 0;
-        if (f4_processRecur(g, _, root) || root->childCount() != 1)
+        if (f4_process(g, root) || root->childCount() != 1)
         {
                 delete root;
@@ -677 +676 @@
 
         // convert genotype strings into tree structures
-        int _1 = 0, _2 = 0;
-        if (f4_processRecur(g1, _1, &root1) || (root1.childCount() != 1)) return GENOPER_OPFAIL;
-        if (f4_processRecur(g2, _2, &root2) || (root2.childCount() != 1)) return GENOPER_OPFAIL;
+        if (f4_process(g1, &root1) || (root1.childCount() != 1)) return GENOPER_OPFAIL;
+        if (f4_process(g2, &root2) || (root2.childCount() != 1)) return GENOPER_OPFAIL;
 
         // decide amounts of crossover, 0.1-0.9

cpp/frams/genetics/f4/f4_oper.h

@@ -72 +72 @@
          * @return GENOOPER_OK if genotype is valid, GENOPER_REPAIR if genotype can be repaired, GENOPER_OPFAIL if genotype can't be repaired
          */
-        int  ValidateRec(f4_Node *geno, int retrycount) const;
+        int  ValidateRecur(f4_Node *geno, int retrycount) const;
 
         /**