polars_io/
shared.rs

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
use std::io::{Read, Write};
use std::sync::Arc;

use arrow::array::new_empty_array;
use arrow::record_batch::RecordBatch;
use polars_core::prelude::*;

use crate::options::RowIndex;
#[cfg(any(feature = "ipc", feature = "avro", feature = "ipc_streaming",))]
use crate::predicates::PhysicalIoExpr;

pub trait SerReader<R>
where
    R: Read,
{
    /// Create a new instance of the [`SerReader`]
    fn new(reader: R) -> Self;

    /// Make sure that all columns are contiguous in memory by
    /// aggregating the chunks into a single array.
    #[must_use]
    fn set_rechunk(self, _rechunk: bool) -> Self
    where
        Self: Sized,
    {
        self
    }

    /// Take the SerReader and return a parsed DataFrame.
    fn finish(self) -> PolarsResult<DataFrame>;
}

pub trait SerWriter<W>
where
    W: Write,
{
    fn new(writer: W) -> Self
    where
        Self: Sized;
    fn finish(&mut self, df: &mut DataFrame) -> PolarsResult<()>;
}

pub trait WriteDataFrameToFile {
    fn write_df_to_file<W: std::io::Write>(&self, df: DataFrame, file: W) -> PolarsResult<()>;
}

pub trait ArrowReader {
    fn next_record_batch(&mut self) -> PolarsResult<Option<RecordBatch>>;
}

#[cfg(any(feature = "ipc", feature = "avro", feature = "ipc_streaming",))]
pub(crate) fn finish_reader<R: ArrowReader>(
    mut reader: R,
    rechunk: bool,
    n_rows: Option<usize>,
    predicate: Option<Arc<dyn PhysicalIoExpr>>,
    arrow_schema: &ArrowSchema,
    row_index: Option<RowIndex>,
) -> PolarsResult<DataFrame> {
    use polars_core::utils::accumulate_dataframes_vertical_unchecked;

    let mut num_rows = 0;
    let mut parsed_dfs = Vec::with_capacity(1024);

    while let Some(batch) = reader.next_record_batch()? {
        let current_num_rows = num_rows as IdxSize;
        num_rows += batch.len();
        let mut df = DataFrame::try_from((batch, arrow_schema))?;

        if let Some(rc) = &row_index {
            df.with_row_index_mut(rc.name.clone(), Some(current_num_rows + rc.offset));
        }

        if let Some(predicate) = &predicate {
            let s = predicate.evaluate_io(&df)?;
            let mask = s.bool().expect("filter predicates was not of type boolean");
            df = df.filter(mask)?;
        }

        if let Some(n) = n_rows {
            if num_rows >= n {
                let len = n - parsed_dfs
                    .iter()
                    .map(|df: &DataFrame| df.height())
                    .sum::<usize>();
                if polars_core::config::verbose() {
                    eprintln!("sliced off {} rows of the 'DataFrame'. These lines were read because they were in a single chunk.", df.height().saturating_sub(n))
                }
                parsed_dfs.push(df.slice(0, len));
                break;
            }
        }
        parsed_dfs.push(df);
    }

    let mut df = {
        if parsed_dfs.is_empty() {
            // Create an empty dataframe with the correct data types
            let empty_cols = arrow_schema
                .iter_values()
                .map(|fld| {
                    Series::try_from((fld.name.clone(), new_empty_array(fld.dtype.clone())))
                        .map(Column::from)
                })
                .collect::<PolarsResult<_>>()?;
            DataFrame::new(empty_cols)?
        } else {
            // If there are any rows, accumulate them into a df
            accumulate_dataframes_vertical_unchecked(parsed_dfs)
        }
    };

    if rechunk {
        df.as_single_chunk_par();
    }
    Ok(df)
}

pub(crate) fn schema_to_arrow_checked(
    schema: &Schema,
    compat_level: CompatLevel,
    _file_name: &str,
) -> PolarsResult<ArrowSchema> {
    schema
        .iter_fields()
        .map(|field| {
            #[cfg(feature = "object")]
            {
                polars_ensure!(
                    !matches!(field.dtype(), DataType::Object(_, _)),
                    ComputeError: "cannot write 'Object' datatype to {}",
                    _file_name
                );
            }

            let field = field
                .dtype()
                .to_arrow_field(field.name().clone(), compat_level);
            Ok((field.name.clone(), field))
        })
        .collect::<PolarsResult<ArrowSchema>>()
}