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
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182

use std::collections::HashMap;
use std::sync::Arc;

use arrow::datatypes::SchemaRef;
use arrow::error::ArrowError;
use arrow::record_batch::{RecordBatch, RecordBatchReader};

use crate::array_decoder::NaiveStripeDecoder;
use crate::error::Result;
use crate::projection::ProjectionMask;
use crate::reader::metadata::{read_metadata, FileMetadata};
use crate::reader::ChunkReader;
use crate::schema::RootDataType;
use crate::stripe::Stripe;

const DEFAULT_BATCH_SIZE: usize = 8192;

pub struct ArrowReaderBuilder<R> {
    pub(crate) reader: R,
    pub(crate) file_metadata: Arc<FileMetadata>,
    pub(crate) batch_size: usize,
    pub(crate) projection: ProjectionMask,
    pub(crate) schema_ref: Option<SchemaRef>,
}

impl<R> ArrowReaderBuilder<R> {
    pub(crate) fn new(reader: R, file_metadata: Arc<FileMetadata>) -> Self {
        Self {
            reader,
            file_metadata,
            batch_size: DEFAULT_BATCH_SIZE,
            projection: ProjectionMask::all(),
            schema_ref: None,
        }
    }

    pub fn file_metadata(&self) -> &FileMetadata {
        &self.file_metadata
    }

    pub fn with_batch_size(mut self, batch_size: usize) -> Self {
        self.batch_size = batch_size;
        self
    }

    pub fn with_projection(mut self, projection: ProjectionMask) -> Self {
        self.projection = projection;
        self
    }

    pub fn with_schema(mut self, schema: SchemaRef) -> Self {
        self.schema_ref = Some(schema);
        self
    }

    /// Returns the currently computed schema
    ///
    /// Unless [`with_schema`](Self::with_schema) was called, this is computed dynamically
    /// based on the current projection and the underlying file format.
    pub fn schema(&self) -> SchemaRef {
        let projected_data_type = self
            .file_metadata
            .root_data_type()
            .project(&self.projection);
        let metadata = self
            .file_metadata
            .user_custom_metadata()
            .iter()
            .map(|(key, value)| (key.clone(), String::from_utf8_lossy(value).to_string()))
            .collect::<HashMap<_, _>>();
        self.schema_ref
            .clone()
            .unwrap_or_else(|| Arc::new(projected_data_type.create_arrow_schema(&metadata)))
    }
}

impl<R: ChunkReader> ArrowReaderBuilder<R> {
    pub fn try_new(mut reader: R) -> Result<Self> {
        let file_metadata = Arc::new(read_metadata(&mut reader)?);
        Ok(Self::new(reader, file_metadata))
    }

    pub fn build(self) -> ArrowReader<R> {
        let schema_ref = self.schema();
        let projected_data_type = self
            .file_metadata
            .root_data_type()
            .project(&self.projection);
        let cursor = Cursor {
            reader: self.reader,
            file_metadata: self.file_metadata,
            projected_data_type,
            stripe_index: 0,
        };
        ArrowReader {
            cursor,
            schema_ref,
            current_stripe: None,
            batch_size: self.batch_size,
        }
    }
}

pub struct ArrowReader<R> {
    cursor: Cursor<R>,
    schema_ref: SchemaRef,
    current_stripe: Option<Box<dyn Iterator<Item = Result<RecordBatch>> + Send>>,
    batch_size: usize,
}

impl<R> ArrowReader<R> {
    pub fn total_row_count(&self) -> u64 {
        self.cursor.file_metadata.number_of_rows()
    }
}

impl<R: ChunkReader> ArrowReader<R> {
    fn try_advance_stripe(&mut self) -> Result<Option<RecordBatch>, ArrowError> {
        let stripe = self.cursor.next().transpose()?;
        match stripe {
            Some(stripe) => {
                let decoder =
                    NaiveStripeDecoder::new(stripe, self.schema_ref.clone(), self.batch_size)?;
                self.current_stripe = Some(Box::new(decoder));
                self.next().transpose()
            }
            None => Ok(None),
        }
    }
}

impl<R: ChunkReader> RecordBatchReader for ArrowReader<R> {
    fn schema(&self) -> SchemaRef {
        self.schema_ref.clone()
    }
}

impl<R: ChunkReader> Iterator for ArrowReader<R> {
    type Item = Result<RecordBatch, ArrowError>;

    fn next(&mut self) -> Option<Self::Item> {
        match self.current_stripe.as_mut() {
            Some(stripe) => {
                match stripe
                    .next()
                    .map(|batch| batch.map_err(|err| ArrowError::ExternalError(Box::new(err))))
                {
                    Some(rb) => Some(rb),
                    None => self.try_advance_stripe().transpose(),
                }
            }
            None => self.try_advance_stripe().transpose(),
        }
    }
}

pub(crate) struct Cursor<R> {
    pub reader: R,
    pub file_metadata: Arc<FileMetadata>,
    pub projected_data_type: RootDataType,
    pub stripe_index: usize,
}

impl<R: ChunkReader> Iterator for Cursor<R> {
    type Item = Result<Stripe>;

    fn next(&mut self) -> Option<Self::Item> {
        self.file_metadata
            .stripe_metadatas()
            .get(self.stripe_index)
            .map(|info| {
                let stripe = Stripe::new(
                    &mut self.reader,
                    &self.file_metadata,
                    &self.projected_data_type.clone(),
                    info,
                );
                self.stripe_index += 1;
                stripe
            })
    }
}