rdblue commented on code in PR #3202: URL: https://github.com/apache/parquet-java/pull/3202#discussion_r2072166725
########## parquet-variant/src/main/java/org/apache/parquet/variant/VariantBuilder.java: ########## @@ -0,0 +1,658 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +package org.apache.parquet.variant; + +import java.math.BigDecimal; +import java.math.BigInteger; +import java.nio.ByteBuffer; +import java.nio.ByteOrder; +import java.nio.charset.StandardCharsets; +import java.util.ArrayList; +import java.util.Arrays; +import java.util.Collections; +import java.util.HashMap; + +/** + * Builder for creating Variant value and metadata. + */ +public class VariantBuilder { + /** The buffer for building the Variant value. The first `writePos` bytes have been written. */ + protected byte[] writeBuffer = new byte[1024]; + + protected int writePos = 0; + /** The dictionary for mapping keys to monotonically increasing ids. */ + private final HashMap<String, Integer> dictionary = new HashMap<>(); + /** The keys in the dictionary, in id order. */ + private final ArrayList<byte[]> dictionaryKeys = new ArrayList<>(); + + /** + * These are used to build nested objects and arrays, via startObject() and startArray(). + * Only one of these can be non-null at a time. If one of these is non-null, then no append() + * methods can be called on this builder, until endObject() or endArray() is called. + */ + protected VariantObjectBuilder objectBuilder = null; + + protected VariantArrayBuilder arrayBuilder = null; + + /** + * Creates a VariantBuilder. + */ + public VariantBuilder() {} + + /** + * @return the Variant value + */ + public Variant build() { + if (objectBuilder != null) { + throw new IllegalStateException( + "Cannot call build() while an object is being built. Must call endObject() first."); + } + if (arrayBuilder != null) { + throw new IllegalStateException( + "Cannot call build() while an array is being built. Must call endArray() first."); + } + int numKeys = dictionaryKeys.size(); + // Use long to avoid overflow in accumulating lengths. + long dictionaryTotalDataSize = 0; + for (byte[] key : dictionaryKeys) { + dictionaryTotalDataSize += key.length; + } + // Determine the number of bytes required per offset entry. + // The largest offset is the one-past-the-end value, which is total data size. It's very + // unlikely that the number of keys could be larger, but incorporate that into the calculation + // in case of pathological data. + long maxSize = Math.max(dictionaryTotalDataSize, numKeys); + int offsetSize = getMinIntegerSize((int) maxSize); + + int offsetListOffset = 1 + offsetSize; + int dataOffset = offsetListOffset + (numKeys + 1) * offsetSize; + long metadataSize = dataOffset + dictionaryTotalDataSize; + + byte[] metadata = new byte[(int) metadataSize]; + // Only unsorted dictionary keys are supported. + // TODO: Support sorted dictionary keys. + int headerByte = VariantUtil.VERSION | ((offsetSize - 1) << 6); + VariantUtil.writeLong(metadata, 0, headerByte, 1); + VariantUtil.writeLong(metadata, 1, numKeys, offsetSize); + int currentOffset = 0; + for (int i = 0; i < numKeys; ++i) { + VariantUtil.writeLong(metadata, offsetListOffset + i * offsetSize, currentOffset, offsetSize); + byte[] key = dictionaryKeys.get(i); + System.arraycopy(key, 0, metadata, dataOffset + currentOffset, key.length); + currentOffset += key.length; + } + VariantUtil.writeLong(metadata, offsetListOffset + numKeys * offsetSize, currentOffset, offsetSize); + // Copying the data to a new buffer, to retain only the required data length, not the capacity. + return new Variant(Arrays.copyOfRange(writeBuffer, 0, writePos), metadata); + } + + /** + * Appends a string value to the Variant builder. + * @param str the string value to append + */ + public void appendString(String str) { + onAppend(); + byte[] data = str.getBytes(StandardCharsets.UTF_8); + boolean longStr = data.length > VariantUtil.MAX_SHORT_STR_SIZE; + checkCapacity((longStr ? 1 + VariantUtil.U32_SIZE : 1) + data.length); + if (longStr) { + writeBuffer[writePos] = VariantUtil.HEADER_LONG_STRING; + writePos += 1; + VariantUtil.writeLong(writeBuffer, writePos, data.length, VariantUtil.U32_SIZE); + writePos += VariantUtil.U32_SIZE; + } else { + writeBuffer[writePos] = VariantUtil.shortStrHeader(data.length); + writePos += 1; + } + System.arraycopy(data, 0, writeBuffer, writePos, data.length); + writePos += data.length; + incrementNumValues(); Review Comment: After looking into `endObject` and `endArray`, I think the rationale for incrementing values at the end of each append is to make the consistency check in `endObject` catch cases where a nested object is partially written. So if you have a nested object, the parent won't have a matching number of keys and values until the inner `endObject` is called. I think we need a stronger validation that doesn't rely on when `incrementNumValues` is called. First, I don't think that check would similarly work when an object in an array is unfinished. So the `validate` methods for both arrays and objects need to call `checkMultipleNested` to verify that there are no unfinished objects or arrays in the child builder. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
