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
183
184
185
186
187
188
189

use smartnoise_validator::errors::*;

use crate::NodeArguments;
use smartnoise_validator::base::{Array, ReleaseNode, Value, IndexKey};
use smartnoise_validator::utilities::{take_argument, get_common_value};
use smartnoise_validator::components::partition::{even_split_lengths, make_dense_partition_keys};
use crate::components::Evaluable;
use ndarray::{ArrayD, Axis};

use smartnoise_validator::{proto, Integer};

use smartnoise_validator::utilities::array::slow_select;
use indexmap::map::IndexMap;


impl Evaluable for proto::Partition {
    fn evaluate(&self, _privacy_definition: &Option<proto::PrivacyDefinition>, mut arguments: NodeArguments) -> Result<ReleaseNode> {
        let data = take_argument(&mut arguments, "data")?;
        Ok(ReleaseNode::new(match arguments.remove::<IndexKey>(&"by".into()) {
            Some(by) => {
                let categories = take_argument(&mut arguments, "categories")?.jagged()?;
                let partitions = make_dense_partition_keys(
                    categories, Some(by.ref_array()?.shape().len() as i64))?;

                match by.array()? {
                    Array::Int(by) =>
                        Value::Partitions(partition_by(&data, by.mapv(IndexKey::from), partitions)?),
                    Array::Bool(by) =>
                        Value::Partitions(partition_by(&data, by.mapv(IndexKey::from), partitions)?),
                    Array::Str(by) =>
                        Value::Partitions(partition_by(&data, by.mapv(IndexKey::from), partitions)?),
                    _ => return Err("by and categories must share the same type".into())
                }
            },
            None => {
                let num_partitions = take_argument(&mut arguments, "num_partitions")?
                    .array()?.first_int()?;

                Value::Partitions(partition_evenly(&data, num_partitions as i64)?)
            }
        }))
    }
}

// to make the nested indexmaps more readable
type ColName = IndexKey;

/// Partitions data evenly into num_partitions partitions
///
/// The first partitions may have one more element than the latter partitions.
///
/// # Arguments
/// * `data` - Data to be partitioned.
/// * `num_partitions` - Number of keys in the indexmap of arrays returned.
///
/// # Return
/// Indexmap with data splits.
///
/// # Example
/// ```
/// use ndarray::{ArrayD, arr1, arr2};
/// use smartnoise_runtime::components::partition::partition_ndarray_evenly;
/// use smartnoise_validator::base::IndexKey;
///
/// let data = arr2(&[ [1, 2], [4, 5], [7, 8], [10, 11] ]).into_dyn();
/// let partitioned = partition_ndarray_evenly(&data, 3);
/// assert_eq!(partitioned.get::<IndexKey>(&0.into()).unwrap().clone(), arr2(&[ [1, 2], [4, 5] ]).into_dyn());
/// assert_eq!(partitioned.get::<IndexKey>(&1.into()).unwrap().clone(), arr2(&[ [7, 8] ]).into_dyn());
/// assert_eq!(partitioned.get::<IndexKey>(&2.into()).unwrap().clone(), arr2(&[ [10, 11] ]).into_dyn());
/// ```
pub fn partition_ndarray_evenly<T: Clone + Default + std::fmt::Debug>(
    data: &ArrayD<T>, num_partitions: i64
) -> IndexMap<IndexKey, ArrayD<T>> {

    let mut offset = 0;
    even_split_lengths(data.len_of(Axis(0)) as i64, num_partitions).into_iter().enumerate()
        .map(|(idx, length)| {

            let entry = (
                IndexKey::from(idx as Integer),
                slow_select(data, Axis(0),
                            &(offset as usize..(offset + length) as usize).collect::<Vec<usize>>())
            );
            offset += length;
            entry
        })
        .collect::<IndexMap<IndexKey, ArrayD<T>>>()
}

pub fn partition_evenly(data: &Value, num_partitions: i64) -> Result<IndexMap<IndexKey, Value>> {
    Ok(match data {
        Value::Dataframe(data) => {

            let columnar_partitions = data.into_iter()
                .map(|(k, v)| Ok((
                    k.clone(),
                    partition_evenly(v, num_partitions)?
                )))
                .collect::<Result<IndexMap<ColName, IndexMap<IndexKey, Value>>>>()?;

            let number_rows: usize = get_common_value(&data.values()
                .map(|v| v.ref_array()?.num_records())
                .collect::<Result<Vec<usize>>>()?)
                .ok_or_else(|| Error::from("columns of a dataframe must share the same length"))?;

            even_split_lengths(number_rows as i64, num_partitions).into_iter()
                .map(|idx| IndexKey::from(idx as Integer))
                .map(|idx| (
                    idx.clone(),
                    Value::Dataframe(columnar_partitions.iter().map(|(colname, partitions)|
                        (colname.clone(), partitions.get(&idx).unwrap().clone())
                    ).collect::<IndexMap<ColName, Value>>())
                ))
                .collect::<IndexMap<IndexKey, Value>>()
        },
        Value::Array(data) => match data {
            Array::Float(data) =>
                partition_ndarray_evenly(data, num_partitions).into_iter()
                    .map(|(idx, data)| (idx, data.into())).collect::<IndexMap<IndexKey, Value>>(),
            Array::Int(data) =>
                partition_ndarray_evenly(data, num_partitions).into_iter()
                    .map(|(idx, data)| (idx, data.into())).collect::<IndexMap<IndexKey, Value>>(),
            Array::Bool(data) =>
                partition_ndarray_evenly(data, num_partitions).into_iter()
                    .map(|(idx, data)| (idx, data.into())).collect::<IndexMap<IndexKey, Value>>(),
            Array::Str(data) =>
                partition_ndarray_evenly(data, num_partitions).into_iter()
                    .map(|(idx, data)| (idx, data.into())).collect::<IndexMap<IndexKey, Value>>(),
        },
        _ => return Err("data: must be a dataframe or array".into())
    })

}

pub fn partition_by(
    data: &Value, by: ArrayD<IndexKey>, partition_keys: Vec<IndexKey>
) -> Result<IndexMap<IndexKey, Value>> {

    let mut indices = partition_keys.into_iter()
        .map(|key| (key, vec![]))
        .collect::<IndexMap<IndexKey, Vec<usize>>>();

    match by.ndim() {
        0 => return Err("by: invalid dimensionality".into()),
        1 => by.into_dimensionality::<ndarray::Ix1>()?,
        _ => by.genrows().into_iter().map(|row| IndexKey::Tuple(row.to_vec())).collect()
    }
        .into_iter().enumerate()
        .for_each(|(idx, cat)| indices.entry(cat.clone())
            .or_insert_with(Vec::new).push(idx));

    // partition either an array or a dataframe
    fn value_partitioner(data: &Value, indices: &IndexMap<IndexKey, Vec<usize>>) -> Result<IndexMap<IndexKey, Value>> {
        Ok(match data {
            Value::Array(data) => match data {
                Array::Int(data) => indices.into_iter()
                    .map(|(cat, idxs)| (cat.clone(), data.select(ndarray::Axis(0), idxs).into()))
                    .collect::<IndexMap<IndexKey, Value>>(),
                Array::Float(data) => indices.into_iter()
                    .map(|(cat, idxs)| (cat.clone(), data.select(ndarray::Axis(0), idxs).into()))
                    .collect::<IndexMap<IndexKey, Value>>(),
                Array::Bool(data) => indices.into_iter()
                    .map(|(cat, idxs)| (cat.clone(), data.select(ndarray::Axis(0), idxs).into()))
                    .collect::<IndexMap<IndexKey, Value>>(),
                Array::Str(data) => indices.into_iter()
                    .map(|(cat, idxs)| (cat.clone(), slow_select(&data, ndarray::Axis(0), idxs).into()))
                    .collect::<IndexMap<IndexKey, Value>>()
            },

            Value::Dataframe(data) => {
                let columnar_partitions = data.into_iter().map(|(k, v)|
                    Ok((k.clone(), value_partitioner(v, indices)?)))
                    .collect::<Result<IndexMap<ColName, IndexMap<IndexKey, Value>>>>()?;

                indices.iter()
                    .map(|(cat, _)| (
                        cat.clone(),
                        Value::Dataframe(columnar_partitions.iter().map(|(colname, partitions)|
                            (colname.clone(), partitions.get(&cat.clone()).unwrap().clone())
                        ).collect::<IndexMap<ColName, Value>>())
                    ))
                    .collect::<IndexMap<IndexKey, Value>>()
            },
            _ => return Err("data: must be a dataframe or array".into())
        })
    };

    value_partitioner(data, &indices)
}