cache result of getArray() (fixes #13)

Author: sbittrich

CHANGELOG.md

@@ -3,6 +3,10 @@ CIFTools Changelog
 
 This project uses semantic versioning. Furthermore, this project provides code that was generated from schemata. Any schema change that introduces a breaking change in the generated code is considered as breaking for the whole project. Additional information is provided below when this occurs (named `Breaking schema changes`). Most of these occur in experimental categories and are unlikely to affect your code. `Breaking API changes` will be avoided starting with version 1.0.0.
 
+ciftools-java 7.0.1 - March 2025
+-------------
+* cache results of `DelegatingColumn#getArray()` to avoid performance penalty if schema type and actual data type differ (fixes #13)
+
 ciftools-java 7.0.0 - March 2025
 -------------
 ### Breaking schema changes

src/main/java/org/rcsb/cif/schema/DelegatingColumn.java

@@ -10,6 +10,7 @@
 public class DelegatingColumn<T> implements Column<T> {
     protected final Column<?> delegate;
     protected final Class<T> type;
+    protected T array;
 
     @SuppressWarnings("unchecked")
     public DelegatingColumn(Column<?> delegate) {
@@ -44,16 +45,12 @@ public ValueKind getValueKind(int row) {
     @SuppressWarnings("unchecked")
     @Override
     public T getArray() {
-        Object array = delegate.getArray();
-        // matches expectation
-        if (type.isInstance(array)) {
-            return (T) array;
-        }
-        // empty column
         if (array == null) {
-            return null;
+            Object raw = delegate.getArray();
+            // 1st condition: type matches expectation, 2nd check: empty column?
+            array = (type.isInstance(raw)) ? (T) raw : (raw != null) ? forceType(raw) : null;
         }
-        return forceType(array);
+        return array;
     }
 
     @SuppressWarnings("unchecked")

src/test/java/org/rcsb/cif/ReaderTest.java

@@ -2,6 +2,7 @@
 
 import org.junit.jupiter.api.Test;
 import org.rcsb.cif.model.CifFile;
+import org.rcsb.cif.model.FloatColumn;
 import org.rcsb.cif.model.IntColumn;
 import org.rcsb.cif.schema.StandardSchemata;
 import org.rcsb.cif.schema.mm.AtomSite;
@@ -10,6 +11,8 @@
 
 import java.io.IOException;
 import java.io.InputStream;
+import java.util.ArrayList;
+import java.util.Comparator;
 import java.util.List;
 import java.util.Map;
 import java.util.OptionalDouble;
@@ -125,4 +128,25 @@ void whenReadingStringWithEmptyQuotation_thenValueAvailable() throws IOException
         String gene = cifFile.getFirstBlock().getCategory("ma_target_ref_db_details").getColumn("gene_name").getStringData(0);
         assertEquals("''cytochrome P450", gene, "Gene name with additional quotes not parsed correctly");
     }
+
+    @Test
+    void whenReadingBinaryColumnWithTypeMismatch_thenPerformanceNotDegrading() throws IOException {
+        String id = "9a2f";
+        InputStream inputStream = TestHelper.getInputStream("bcif/" + id + ".bcif.gz");
+        AtomSite atomSite = CifIO.readFromInputStream(inputStream).as(StandardSchemata.MMCIF).getFirstBlock().getAtomSite();
+
+        // schema type is float but data is internally encoded as int -- assert that this doesn't degrade performance when accessing values one-by-one
+        FloatColumn bIsoOrEquiv = atomSite.getBIsoOrEquiv();
+
+        long start = System.nanoTime();
+        double sum = 0.0;
+        for (int i = 0; i < bIsoOrEquiv.getRowCount(); i++) {
+            // invokes `getArray()`, which used to convert the internal array representation again and again if types diverged
+            sum += bIsoOrEquiv.get(i);
+        }
+        assertEquals(0.0, sum);
+        long end = System.nanoTime();
+        long delta_ms = (end - start) / 1_000_000;
+        assertTrue(delta_ms < 250, "Access to  took " + delta_ms + " ms and we deem that too slow");
+    }
 }