better reproduction of compact newick files (removal of unecessary spaces)

[jalview.git] / src / jalview / io / NewickFile.java

diff --git a/src/jalview/io/NewickFile.java b/src/jalview/io/NewickFile.java
index 4c5de3a..aa9a129 100755 (executable)
--- a/src/jalview/io/NewickFile.java
+++ b/src/jalview/io/NewickFile.java
@@ -30,9 +30,9 @@ import java.io.*;
 import jalview.datamodel.*;
 
 /**
- * DOCUMENT ME!
- *
- * @author $author$
+ * Parse a new hanpshire style tree
+ * Caveats: NHX files are NOT supported and the tree distances and topology are unreliable when they are parsed.
+ * @author Jim Procter
  * @version $Revision$
  */
 public class NewickFile
@@ -45,7 +45,7 @@ public class NewickFile
 
   // File IO Flags
   boolean ReplaceUnderscores = false;
-  boolean printRootInfo = false;
+  boolean printRootInfo = true;
   private com.stevesoft.pat.Regex[] NodeSafeName = new com.stevesoft.pat.Regex[]
       {
       new com.stevesoft.pat.Regex().perlCode("m/[\\[,:'()]/"), // test for requiring quotes
@@ -211,7 +211,7 @@ public class NewickFile
     String nodename = null;
 
     float DefDistance = (float) 0.001; // @param Default distance for a node - very very small
-    int DefBootstrap = 0; // @param Default bootstrap for a node
+    int DefBootstrap = -1; // @param Default bootstrap for a node
 
     float distance = DefDistance;
     int bootstrap = DefBootstrap;
@@ -221,32 +221,14 @@ public class NewickFile
     com.stevesoft.pat.Regex majorsyms = new com.stevesoft.pat.Regex(
         "[(\\['),;]");
 
+    int nextcp=0;
+    int ncp = cp;
     while (majorsyms.searchFrom(nf, cp) && (Error == null))
     {
       int fcp = majorsyms.matchedFrom();
-
-      switch (nf.charAt(fcp))
+      char schar;
+      switch (schar=nf.charAt(fcp))
       {
-        case '[': // Comment or structured/extended NH format info
-
-          com.stevesoft.pat.Regex comment = new com.stevesoft.pat.Regex(
-              "]");
-
-          if (comment.searchFrom(nf, fcp))
-          {
-            // Skip the comment field
-            cp = 1 + comment.matchedFrom();
-          }
-          else
-          {
-            Error = ErrorStringrange(Error, "Unterminated comment", 3,
-                                     fcp, nf);
-          }
-
-          ;
-
-          break;
-
         case '(':
 
           // ascending should not be set
@@ -317,24 +299,61 @@ public class NewickFile
 
           break;
 
-        case ';':
-
-          if (d != -1)
+        default:
+          if (schar==';')
           {
-            Error = ErrorStringrange(Error,
+            if (d != -1)
+            {
+              Error = ErrorStringrange(Error,
                                      "Wayward semicolon (depth=" + d + ")", 7,
                                      fcp, nf);
+            }
+            // cp advanced at the end of default
           }
+          if (schar == '[')
+          { 
+            // node string contains Comment or structured/extended NH format info
+            /* if ((fcp-cp>1 && nf.substring(cp,fcp).trim().length()>1))
+              {
+                // will process in remains System.err.println("skipped text: '"+nf.substring(cp,fcp)+"'");
+              }
+             */
+            // verify termination.
+            com.stevesoft.pat.Regex comment = new com.stevesoft.pat.Regex(
+              "]");
+            if (comment.searchFrom(nf, fcp))
+            {
+              // Skip the comment field
+              nextcp=comment.matchedFrom()+1;
+              warningMessage = "Tree file contained comments which may confuse input algorithm.";
+              break;
+              
+              // cp advanced at the end of default to nextcp, ncp is unchanged so any node info can be read.
+            }
+            else
+            {
+              Error = ErrorStringrange(Error, "Unterminated comment", 3,
+                                     fcp, nf);
+            }
 
-          // cp advanced at the end of default
-        default:
-
+            ;
+          }
           // Parse simpler field strings
-          String fstring = nf.substring(cp, fcp);
+          String fstring = nf.substring(ncp, fcp);
+          // remove any comments before we parse the node info
+          // TODO: test newick file with quoted square brackets in node name (is this allowed?)
+          while (fstring.indexOf(']')>-1)
+          {
+            int cstart=fstring.indexOf('[');
+            int cend=fstring.indexOf(']');
+            String comment =  fstring.substring(cstart+1,cend);
+            fstring = fstring.substring(0, cstart)+fstring.substring(cend+1);
+            
+          }
           com.stevesoft.pat.Regex uqnodename = new com.stevesoft.pat.Regex(
               "\\b([^' :;\\](),]+)");
           com.stevesoft.pat.Regex nbootstrap = new com.stevesoft.pat.Regex(
-              "\\S+([0-9+]+)\\S*:");
+              "\\s*([0-9+]+)\\s*:");
           com.stevesoft.pat.Regex ndist = new com.stevesoft.pat.Regex(
               ":([-0-9Ee.+]+)");
 
@@ -362,9 +381,14 @@ public class NewickFile
             }
           }
 
-          if (nbootstrap.search(fstring) &&
-              (nbootstrap.matchedFrom(1) > (uqnodename.matchedFrom(1) +
-                                            uqnodename.stringMatched().length())))
+          if (nbootstrap.search(fstring))
+            {
+            if (nbootstrap.stringMatched(1).equals(uqnodename.stringMatched(1)))
+              {
+                nodename=null; // no nodename here.
+              }
+            if (nodename==null || nodename.length()==0 || nbootstrap.matchedFrom(1) > (uqnodename.matchedFrom(1) + 
+                      uqnodename.stringMatched().length()))
           {
             try
             {
@@ -375,9 +399,10 @@ public class NewickFile
             {
               Error = ErrorStringrange(Error,
                                        "Can't parse bootstrap value", 4,
-                                       cp + nbootstrap.matchedFrom(), nf);
+                                       ncp + nbootstrap.matchedFrom(), nf);
             }
           }
+            }
 
           boolean nodehasdistance = false;
 
@@ -393,7 +418,7 @@ public class NewickFile
             {
               Error = ErrorStringrange(Error,
                                        "Can't parse node distance value", 7,
-                                       cp + ndist.matchedFrom(), nf);
+                                       ncp + ndist.matchedFrom(), nf);
             }
           }
 
@@ -474,16 +499,20 @@ public class NewickFile
                 }
               }
             }
-
-            // else : We do nothing if ';' is encountered.
           }
 
           // Reset new node properties to obvious fakes
           nodename = null;
           distance = DefDistance;
           bootstrap = DefBootstrap;
-
-          cp = fcp + 1;
+      }
+      if (nextcp==0)
+      {
+        ncp = cp = fcp + 1;
+      }
+      else {
+        cp=nextcp;
+        nextcp=0;
       }
     }
 
@@ -658,7 +687,7 @@ public class NewickFile
   {
     return ( (c.getName() == null) ? "" : nodeName(c.getName())) +
         ( (HasBootstrap)
-         ? ( (c.getBootstrap() > -1) ? (" " + c.getBootstrap()) : "") : "") +
+         ? ( (c.getBootstrap() > -1) ? ((c.getName()!=null ? " " : "")+ c.getBootstrap()) : "") : "") +
         ( (HasDistances) ? (":" + c.dist) : "");
   }
 
@@ -674,7 +703,8 @@ public class NewickFile
     return (printRootInfo)
         ? ( ( (root.getName() == null) ? "" : nodeName(root.getName())) +
            ( (HasBootstrap)
-            ? ( (root.getBootstrap() > -1) ? (" " + root.getBootstrap()) : "") :
+            ? ( (root.getBootstrap() > -1) ? ((root.getName()!=null ? " " : "")+
+                    + root.getBootstrap()) : "") :
             "") +
            ( (RootHasDistance) ? (":" + root.dist) : "")) : "";
   }