Issue #517: Restore QbeastUtil.computeHistogramForColumn method (#521)

src/main/scala/io/qbeast/utils/QbeastUtils.scala

@@ -217,4 +217,56 @@ object QbeastUtils extends Logging {
 
   }
 
+  /**
+   * Compute the histogram for a given column
+   *
+   * Since computing the histogram can be expensive, this method is used outside the indexing
+   * process.
+   *
+   * It outputs the histogram of the column as format [bin1, bin2, bin3, ...] Number of bins by
+   * default is 50
+   *
+   * For example:
+   *
+   * val qbeastTable = QbeastTable.forPath(spark, "path")
+   *
+   * val histogram =qbeastTable.computeHistogramForColumn(df, "column")
+   *
+   * df.write.format("qbeast").option("columnsToIndex",
+   * "column:histogram").option("columnStats",histogram).save()
+   *
+   * @param df
+   * @param columnName
+   */
+  @deprecated("Use computeQuantilesForColumn method instead", "0.8.0")
+  def computeHistogramForColumn(df: DataFrame, columnName: String, numBins: Int = 50): String = {
+
+    import df.sparkSession.implicits._
+    if (!df.columns.contains(columnName)) {
+      throw AnalysisExceptionFactory.create(s"Column $columnName does not exist in the dataframe")
+    }
+
+    log.info(s"Computing histogram for column $columnName with number of bins $numBins")
+    val binStarts = "__bin_starts"
+    val stringPartitionColumn =
+      MultiDimClusteringFunctions.range_partition_id(col(columnName), numBins)
+
+    val histogram = df
+      .select(columnName)
+      .distinct()
+      .na
+      .drop()
+      .groupBy(stringPartitionColumn)
+      .agg(min(columnName).alias(binStarts))
+      .select(binStarts)
+      .orderBy(binStarts)
+      .as[String]
+      .collect()
+
+    log.info(s"Histogram for column $columnName: $histogram")
+    histogram
+      .map(string => s"'$string'")
+      .mkString("[", ",", "]")
+  }
+
 }