Add CSV Reader options classes to pylibcudf (#17412)

This PR adds the CSV reader options classes to pylibcudf and plumbs the changes through cudf python.

Authors:
  - Matthew Murray (https://github.com/Matt711)

Approvers:
  - Matthew Roeschke (https://github.com/mroeschke)

URL: #17412

python/cudf/cudf/_lib/csv.pyx

@@ -202,46 +202,71 @@ def read_csv(
             raise ValueError(
                 "dtype should be a scalar/str/list-like/dict-like"
             )
+    options = (
+        plc.io.csv.CsvReaderOptions.builder(plc.io.SourceInfo([datasource]))
+        .compression(c_compression)
+        .mangle_dupe_cols(mangle_dupe_cols)
+        .byte_range_offset(byte_range[0])
+        .byte_range_size(byte_range[1])
+        .nrows(nrows if nrows is not None else -1)
+        .skiprows(skiprows)
+        .skipfooter(skipfooter)
+        .quoting(quoting)
+        .lineterminator(str(lineterminator))
+        .quotechar(quotechar)
+        .decimal(decimal)
+        .delim_whitespace(delim_whitespace)
+        .skipinitialspace(skipinitialspace)
+        .skip_blank_lines(skip_blank_lines)
+        .doublequote(doublequote)
+        .keep_default_na(keep_default_na)
+        .na_filter(na_filter)
+        .dayfirst(dayfirst)
+        .build()
+    )
+
+    options.set_header(header)
+
+    if names is not None:
+        options.set_names([str(name) for name in names])
+
+    if prefix is not None:
+        options.set_prefix(prefix)
+
+    if usecols is not None:
+        if all(isinstance(col, int) for col in usecols):
+            options.set_use_cols_indexes(list(usecols))
+        else:
+            options.set_use_cols_names([str(name) for name in usecols])
+
+    if delimiter is not None:
+        options.set_delimiter(delimiter)
+
+    if thousands is not None:
+        options.set_thousands(thousands)
 
-    lineterminator = str(lineterminator)
+    if comment is not None:
+        options.set_comment(comment)
+
+    if parse_dates is not None:
+        options.set_parse_dates(list(parse_dates))
+
+    if hex_cols is not None:
+        options.set_parse_hex(list(hex_cols))
+
+    options.set_dtypes(new_dtypes)
+
+    if true_values is not None:
+        options.set_true_values([str(val) for val in true_values])
+
+    if false_values is not None:
+        options.set_false_values([str(val) for val in false_values])
+
+    if na_values is not None:
+        options.set_na_values([str(val) for val in na_values])
 
     df = cudf.DataFrame._from_data(
-        *data_from_pylibcudf_io(
-            plc.io.csv.read_csv(
-                plc.io.SourceInfo([datasource]),
-                lineterminator=lineterminator,
-                quotechar = quotechar,
-                quoting = quoting,
-                doublequote = doublequote,
-                header = header,
-                mangle_dupe_cols = mangle_dupe_cols,
-                usecols = usecols,
-                delimiter = delimiter,
-                delim_whitespace = delim_whitespace,
-                skipinitialspace = skipinitialspace,
-                col_names = names,
-                dtypes = new_dtypes,
-                skipfooter = skipfooter,
-                skiprows = skiprows,
-                dayfirst = dayfirst,
-                compression = c_compression,
-                thousands = thousands,
-                decimal = decimal,
-                true_values = true_values,
-                false_values = false_values,
-                nrows = nrows if nrows is not None else -1,
-                byte_range_offset = byte_range[0],
-                byte_range_size = byte_range[1],
-                skip_blank_lines = skip_blank_lines,
-                parse_dates = parse_dates,
-                parse_hex = hex_cols,
-                comment = comment,
-                na_values = na_values,
-                keep_default_na = keep_default_na,
-                na_filter = na_filter,
-                prefix = prefix,
-            )
-        )
+        *data_from_pylibcudf_io(plc.io.csv.read_csv(options))
     )
 
     if dtype is not None:

python/cudf_polars/cudf_polars/dsl/ir.py

@@ -476,23 +476,28 @@ def do_evaluate(
                 with path.open() as f:
                     while f.readline() == "\n":
                         skiprows += 1
-                tbl_w_meta = plc.io.csv.read_csv(
-                    plc.io.SourceInfo([path]),
-                    delimiter=sep,
-                    quotechar=quote,
-                    lineterminator=eol,
-                    col_names=column_names,
-                    header=header,
-                    usecols=usecols,
-                    na_filter=True,
-                    na_values=null_values,
-                    keep_default_na=False,
-                    skiprows=skiprows,
-                    comment=comment,
-                    decimal=decimal,
-                    dtypes=schema,
-                    nrows=n_rows,
+                options = (
+                    plc.io.csv.CsvReaderOptions.builder(plc.io.SourceInfo([path]))
+                    .nrows(n_rows)
+                    .skiprows(skiprows)
+                    .lineterminator(str(eol))
+                    .quotechar(str(quote))
+                    .decimal(decimal)
+                    .keep_default_na(keep_default_na=False)
+                    .na_filter(na_filter=True)
+                    .build()
                 )
+                options.set_delimiter(str(sep))
+                if column_names is not None:
+                    options.set_names([str(name) for name in column_names])
+                options.set_header(header)
+                options.set_dtypes(schema)
+                if usecols is not None:
+                    options.set_use_cols_names([str(name) for name in usecols])
+                options.set_na_values(null_values)
+                if comment is not None:
+                    options.set_comment(comment)
+                tbl_w_meta = plc.io.csv.read_csv(options)
                 pieces.append(tbl_w_meta)
                 if read_partial:
                     n_rows -= tbl_w_meta.tbl.num_rows()

python/pylibcudf/pylibcudf/io/csv.pxd

@@ -6,11 +6,63 @@ from libcpp cimport bool
 from pylibcudf.libcudf.io.csv cimport (
     csv_writer_options,
     csv_writer_options_builder,
+    csv_reader_options,
+    csv_reader_options_builder,
 )
-from pylibcudf.libcudf.io.types cimport quote_style
-from pylibcudf.io.types cimport SinkInfo
+from pylibcudf.io.types cimport SinkInfo, SourceInfo, TableWithMetadata
 from pylibcudf.table cimport Table
 
+from pylibcudf.libcudf.io.types cimport (
+    compression_type,
+    quote_style,
+    table_with_metadata,
+)
+from pylibcudf.libcudf.types cimport size_type
+
+cdef class CsvReaderOptions:
+    cdef csv_reader_options c_obj
+    cdef SourceInfo source
+    cpdef void set_header(self, size_type header)
+    cpdef void set_names(self, list col_names)
+    cpdef void set_prefix(self, str prefix)
+    cpdef void set_use_cols_indexes(self, list col_indices)
+    cpdef void set_use_cols_names(self, list col_names)
+    cpdef void set_delimiter(self, str delimiter)
+    cpdef void set_thousands(self, str thousands)
+    cpdef void set_comment(self, str comment)
+    cpdef void set_parse_dates(self, list val)
+    cpdef void set_parse_hex(self, list val)
+    cpdef void set_dtypes(self, object types)
+    cpdef void set_true_values(self, list true_values)
+    cpdef void set_false_values(self, list false_values)
+    cpdef void set_na_values(self, list na_values)
+
+
+cdef class CsvReaderOptionsBuilder:
+    cdef csv_reader_options_builder c_obj
+    cdef SourceInfo source
+    cpdef CsvReaderOptionsBuilder compression(self, compression_type compression)
+    cpdef CsvReaderOptionsBuilder mangle_dupe_cols(self, bool mangle_dupe_cols)
+    cpdef CsvReaderOptionsBuilder byte_range_offset(self, size_t byte_range_offset)
+    cpdef CsvReaderOptionsBuilder byte_range_size(self, size_t byte_range_size)
+    cpdef CsvReaderOptionsBuilder nrows(self, size_type nrows)
+    cpdef CsvReaderOptionsBuilder skiprows(self, size_type skiprows)
+    cpdef CsvReaderOptionsBuilder skipfooter(self, size_type skipfooter)
+    cpdef CsvReaderOptionsBuilder quoting(self, quote_style quoting)
+    cpdef CsvReaderOptionsBuilder lineterminator(self, str lineterminator)
+    cpdef CsvReaderOptionsBuilder quotechar(self, str quotechar)
+    cpdef CsvReaderOptionsBuilder decimal(self, str decimal)
+    cpdef CsvReaderOptionsBuilder delim_whitespace(self, bool delim_whitespace)
+    cpdef CsvReaderOptionsBuilder skipinitialspace(self, bool skipinitialspace)
+    cpdef CsvReaderOptionsBuilder skip_blank_lines(self, bool skip_blank_lines)
+    cpdef CsvReaderOptionsBuilder doublequote(self, bool doublequote)
+    cpdef CsvReaderOptionsBuilder keep_default_na(self, bool keep_default_na)
+    cpdef CsvReaderOptionsBuilder na_filter(self, bool na_filter)
+    cpdef CsvReaderOptionsBuilder dayfirst(self, bool dayfirst)
+    cpdef CsvReaderOptions build(self)
+
+cpdef TableWithMetadata read_csv(CsvReaderOptions options)
+
 cdef class CsvWriterOptions:
     cdef csv_writer_options c_obj
     cdef Table table

python/pylibcudf/pylibcudf/io/csv.pyi

@@ -2,6 +2,8 @@
 
 from collections.abc import Mapping
 
+from typing_extensions import Self
+
 from pylibcudf.io.types import (
     CompressionType,
     QuoteStyle,
@@ -12,6 +14,47 @@ from pylibcudf.io.types import (
 from pylibcudf.table import Table
 from pylibcudf.types import DataType
 
+class CsvReaderOptions:
+    def __init__(self): ...
+    def set_header(self, header: int): ...
+    def set_names(self, col_names: list[str]): ...
+    def set_prefix(self, prefix: str): ...
+    def set_use_cols_indexes(self, col_indices: list[int]): ...
+    def set_use_cols_names(self, col_names: list[str]): ...
+    def set_delimiter(self, delimiter: str): ...
+    def set_thousands(self, thousands: str): ...
+    def set_comment(self, comment: str): ...
+    def set_parse_dates(self, val: list[int | str]): ...
+    def set_parse_hex(self, val: list[int | str]): ...
+    def set_dtypes(self, types: dict[str, DataType] | list[DataType]): ...
+    def set_true_values(self, true_values: list[str]): ...
+    def set_false_values(self, false_values: list[str]): ...
+    def set_na_values(self, na_values: list[str]): ...
+    @staticmethod
+    def builder(source: SourceInfo) -> CsvReaderOptionsBuilder: ...
+
+class CsvReaderOptionsBuilder:
+    def __init__(self): ...
+    def compression(self, compression: CompressionType) -> Self: ...
+    def mangle_dupe_cols(self, mangle_dupe_cols: bool) -> Self: ...
+    def byte_range_offset(self, byte_range_offset: int) -> Self: ...
+    def byte_range_size(self, byte_range_size: int) -> Self: ...
+    def nrows(self, nrows: int) -> Self: ...
+    def skiprows(self, skiprows: int) -> Self: ...
+    def skipfooter(self, skipfooter: int) -> Self: ...
+    def quoting(self, quoting: QuoteStyle) -> Self: ...
+    def lineterminator(self, lineterminator: str) -> Self: ...
+    def quotechar(self, quotechar: str) -> Self: ...
+    def decimal(self, decimal: str) -> Self: ...
+    def delim_whitespace(self, delim_whitespace: bool) -> Self: ...
+    def skipinitialspace(self, skipinitialspace: bool) -> Self: ...
+    def skip_blank_lines(self, skip_blank_lines: bool) -> Self: ...
+    def doublequote(self, doublequote: bool) -> Self: ...
+    def keep_default_na(self, keep_default_na: bool) -> Self: ...
+    def na_filter(self, na_filter: bool) -> Self: ...
+    def dayfirst(self, dayfirst: bool) -> Self: ...
+    def build(self) -> CsvReaderOptions: ...
+
 def read_csv(
     source_info: SourceInfo,
     *,
@@ -54,7 +97,7 @@ def read_csv(
     # detect_whitespace_around_quotes: bool = False,
     # timestamp_type: DataType = DataType(type_id.EMPTY),
 ) -> TableWithMetadata: ...
-def write_csv(options: CsvWriterOptionsBuilder) -> None: ...
+def write_csv(options: CsvWriterOptionsBuilder): ...
 
 class CsvWriterOptions:
     def __init__(self): ...
@@ -63,14 +106,12 @@ class CsvWriterOptions:
 
 class CsvWriterOptionsBuilder:
     def __init__(self): ...
-    def names(self, names: list) -> CsvWriterOptionsBuilder: ...
-    def na_rep(self, val: str) -> CsvWriterOptionsBuilder: ...
-    def include_header(self, val: bool) -> CsvWriterOptionsBuilder: ...
-    def rows_per_chunk(self, val: int) -> CsvWriterOptionsBuilder: ...
-    def line_terminator(self, term: str) -> CsvWriterOptionsBuilder: ...
-    def inter_column_delimiter(
-        self, delim: str
-    ) -> CsvWriterOptionsBuilder: ...
-    def true_value(self, val: str) -> CsvWriterOptionsBuilder: ...
-    def false_value(self, val: str) -> CsvWriterOptionsBuilder: ...
+    def names(self, names: list) -> Self: ...
+    def na_rep(self, val: str) -> Self: ...
+    def include_header(self, val: bool) -> Self: ...
+    def rows_per_chunk(self, val: int) -> Self: ...
+    def line_terminator(self, term: str) -> Self: ...
+    def inter_column_delimiter(self, delim: str) -> Self: ...
+    def true_value(self, val: str) -> Self: ...
+    def false_value(self, val: str) -> Self: ...
     def build(self) -> CsvWriterOptions: ...