arrow-ipc/src/writer.rs (4 lines): - line 250: // TODO: custom metadata - line 280: // TODO: custom metadata - line 1055: meta as i32, // TODO: is this still applicable? - line 2621: // TODO test file writer/reader parquet/src/encodings/decoding.rs (4 lines): - line 206: // TODO: check validity of the input arguments? - line 843: // TODO: add memory tracker to this - line 972: // TODO: add memory tracker to this - line 1044: // TODO: this is awkward - maybe we should add a non-vectorized API? parquet/src/file/serialized_reader.rs (4 lines): - line 315: // TODO: fix PARQUET-816 - line 467: // TODO: page header could be huge because of statistics. We should set a - line 1206: // TODO: test for every column - line 1300: // TODO: test for every column parquet/src/arrow/arrow_writer/levels.rs (3 lines): - line 314: // TODO: Faster bitmask iteration (#1757) - line 375: // TODO: Faster bitmask iteration (#1757) - line 502: // TODO: Faster bitmask iteration (#1757) parquet/src/arrow/schema/mod.rs (3 lines): - line 185: // TODO: change after addressing ARROW-9777 - line 2232: // TODO: roundtrip - line 2253: // TODO: roundtrip parquet/src/schema/types.rs (2 lines): - line 105: // TODO: should we return `&[&Type]` here? - line 1406: // TODO: add tests for v2 types parquet/src/encodings/encoding/mod.rs (2 lines): - line 56: // TODO: this is pretty inefficient. Revisit in future. - line 738: // TODO: investigate if we can merge lengths and suffixes arrow-csv/src/reader/mod.rs (2 lines): - line 938: // TODO: we should surface the underlying error here. - line 1026: // TODO: we should surface the underlying error here. arrow-schema/src/fields.rs (2 lines): - line 403: // TODO: This currently may produce duplicate type IDs (#3982) - line 434: // TODO: Should this validate type IDs are unique (#3982) parquet/src/file/statistics.rs (2 lines): - line 130: // TODO this should be `None` if there is no information about NULLS. - line 1149: // TODO: remove branch when we no longer support assuming null_count==None in the thrift arrow-flight/src/encode.rs (2 lines): - line 646: /// TODO: at allow dictionaries to be flushed / avoid building them - line 648: /// TODO limit on the number of dictionaries??? parquet/src/data_type.rs (2 lines): - line 596: // TODO - Can Int96 and bool be implemented in these terms? - line 914: // TODO - Remove the duplication between this and the general slice method arrow-ipc/src/reader.rs (2 lines): - line 133: // TODO investigate whether just knowing the number of buffers could - line 2018: // TODO: equality not yet implemented for union, so we check that the length of the array is arrow-array/src/array/union_array.rs (2 lines): - line 831: // TODO: bench on avx512f(feature is still unstable) - line 840: // TODO: bench on non x86 parquet/src/column/reader/decoder.rs (2 lines): - line 234: // TODO: Push vec into decoder (#5177) - line 319: // TODO: Push vec into decoder (#5177) parquet_derive/src/parquet_field.rs (2 lines): - line 293: // TODO: Support group types - line 294: // TODO: Add length if dealing with fixedlenbinary arrow-integration-test/src/lib.rs (2 lines): - line 130: dictionary: None, // TODO: not enough info - line 305: /// TODO: convert to use an Into arrow-select/src/take.rs (2 lines): - line 585: // TODO: Some optimizations can be done here such as if it is - line 768: // TODO: benchmark this function, there might be a faster unsafe alternative parquet/src/encodings/rle.rs (1 line): - line 48: // TODO: tracking memory usage parquet/src/arrow/async_reader/mod.rs (1 line): - line 570: // TODO: calling build_array multiple times is wasteful parquet/src/arrow/array_reader/mod.rs (1 line): - line 64: // TODO: this function is never used, and the trait is not public. Perhaps this should be arrow-schema/src/field.rs (1 line): - line 816: // TODO: improve display with crate https://crates.io/crates/derive_more ? parquet/src/arrow/array_reader/byte_array_dictionary.rs (1 line): - line 334: // TODO: Push vec into decoder (#5177) arrow-json/src/reader/struct_array.rs (1 line): - line 78: // TODO: Investigate how to extract duplicated logic. parquet/src/schema/printer.rs (1 line): - line 145: // TODO: better if we can pass fmt::Write to Printer. dev/release/label_issues.py (1 line): - line 101: # TODO: Find all tickets merged since this tag parquet/src/basic.rs (1 line): - line 564: // TODO: Should this take converted and logical type, for compatibility? arrow-schema/src/extension/canonical/variable_shape_tensor.rs (1 line): - line 98: // TODO: are all data types are suitable as value type? arrow-data/src/equal/list.rs (1 line): - line 118: // TODO: Could potentially compare runs of not NULL values parquet/src/column/page.rs (1 line): - line 210: // TODO: Add support for crc checksum arrow-data/src/data.rs (1 line): - line 1328: // TODO file follow on ticket for full union validation arrow-flight/src/sql/client.rs (1 line): - line 508: // TODO: consume self instead of &mut self to explicitly prevent reuse? parquet/src/util/bit_util.rs (1 line): - line 441: // TODO: better to avoid copying here arrow-schema/src/extension/canonical/fixed_shape_tensor.rs (1 line): - line 94: // TODO: are all data types are suitable as value type? arrow-integration-testing/src/lib.rs (1 line): - line 132: // TODO: convert to a concrete Arrow type parquet/src/arrow/array_reader/fixed_len_byte_array.rs (1 line): - line 166: // TODO: An improvement might be to do this conversion on read parquet/src/file/page_index/offset_index.rs (1 line): - line 54: // TODO: remove annotation after merge parquet/src/arrow/schema/primitive.rs (1 line): - line 303: // TODO: This should check the type length for the decimal and interval types arrow-array/src/builder/map_builder.rs (1 line): - line 240: let array_data = ArrayData::builder(DataType::Map(map_field, false)) // TODO: support sorted keys arrow-schema/src/datatype_parse.rs (1 line): - line 683: // TODO support more structured types (List, LargeList, Struct, Union, Map, RunEndEncoded, etc) arrow-cast/src/cast/string.rs (1 line): - line 49: // TODO: replace with write to builder after https://github.com/apache/arrow-rs/issues/6373 parquet/src/file/page_index/index.rs (1 line): - line 200: // TODO: should we assert (hist.len() % len) == 0? parquet/src/column/writer/mod.rs (1 line): - line 460: // TODO: find out why we don't account for size of levels when we estimate page arrow-array/src/array/primitive_array.rs (1 line): - line 1433: // TODO: the macro is needed here because we'd get "conflicting implementations" error parquet/src/column/reader.rs (1 line): - line 544: // TODO: should we return false if read_new_page() = true and parquet/src/encryption/decrypt.rs (1 line): - line 146: // TODO: In GCM-CTR mode will this need to be a non-GCM decryptor? arrow-array/src/record_batch.rs (1 line): - line 695: // TODO: implement `TryFrom` trait, once parquet/src/file/metadata/reader.rs (1 line): - line 340: // FIXME: there are differing implementations in the case where page indexes are missing arrow-array/src/array/byte_view_array.rs (1 line): - line 703: // TODO: the original buffer could still be used