|
| 1 | +use std::collections::{BTreeSet, VecDeque}; |
| 2 | +use std::mem; |
| 3 | +use std::sync::atomic::{AtomicBool, Ordering}; |
| 4 | + |
| 5 | +use itertools::Itertools; |
| 6 | +use vortex_array::stats::ArrayStatistics; |
| 7 | +use vortex_error::{vortex_bail, VortexExpect, VortexResult}; |
| 8 | + |
| 9 | +use crate::file::{BatchRead, LayoutReader, MessageLocator, RowMask}; |
| 10 | + |
| 11 | +pub enum SplitMask { |
| 12 | + ReadMore(Vec<MessageLocator>), |
| 13 | + Mask(RowMask), |
| 14 | +} |
| 15 | + |
| 16 | +enum SplitState { |
| 17 | + Ranges(Box<dyn Iterator<Item = (usize, usize)> + Send>), |
| 18 | + Splits(BTreeSet<usize>), |
| 19 | +} |
| 20 | + |
| 21 | +/// Iterator over row ranges of a vortex file with bitmaps of valid values in those ranges |
| 22 | +pub trait MaskIterator: Iterator<Item = VortexResult<SplitMask>> + Send { |
| 23 | + /// Register additional horizontal row boundaries to split the generated layout on |
| 24 | + fn additional_splits(&mut self, splits: &mut BTreeSet<usize>) -> VortexResult<()>; |
| 25 | +} |
| 26 | + |
| 27 | +/// MaskIterator that reads boolean arrays out of provided reader and further filters generated masks |
| 28 | +/// |
| 29 | +/// Arrays returned by the reader must be of boolean dtype. |
| 30 | +pub struct FilteringRowSplitIterator { |
| 31 | + reader: Box<dyn LayoutReader>, |
| 32 | + static_splits: FixedSplitIterator, |
| 33 | + in_progress_masks: VecDeque<RowMask>, |
| 34 | + registered_splits: AtomicBool, |
| 35 | +} |
| 36 | + |
| 37 | +impl FilteringRowSplitIterator { |
| 38 | + pub fn new(reader: Box<dyn LayoutReader>, row_count: u64, row_mask: Option<RowMask>) -> Self { |
| 39 | + let static_splits = FixedSplitIterator::new(row_count, row_mask); |
| 40 | + Self { |
| 41 | + reader, |
| 42 | + static_splits, |
| 43 | + in_progress_masks: VecDeque::new(), |
| 44 | + registered_splits: AtomicBool::new(false), |
| 45 | + } |
| 46 | + } |
| 47 | + |
| 48 | + /// Read given mask out of the reader |
| 49 | + fn read_mask(&mut self, mask: RowMask) -> VortexResult<Option<SplitMask>> { |
| 50 | + if let Some(rs) = self.reader.read_selection(&mask)? { |
| 51 | + return match rs { |
| 52 | + BatchRead::ReadMore(rm) => { |
| 53 | + // If the reader needs more data we put the mask back into queue for to come back to it later |
| 54 | + self.in_progress_masks.push_back(mask); |
| 55 | + Ok(Some(SplitMask::ReadMore(rm))) |
| 56 | + } |
| 57 | + BatchRead::Batch(batch) => { |
| 58 | + // If the mask is all FALSE we can safely discard it |
| 59 | + if batch |
| 60 | + .statistics() |
| 61 | + .compute_true_count() |
| 62 | + .vortex_expect("must be a bool array if it's a result of a filter") |
| 63 | + == 0 |
| 64 | + { |
| 65 | + return Ok(None); |
| 66 | + } |
| 67 | + // Combine requested mask with the result of filter read |
| 68 | + Ok(Some(SplitMask::Mask(mask.and_bitmask(batch)?))) |
| 69 | + } |
| 70 | + }; |
| 71 | + } |
| 72 | + Ok(None) |
| 73 | + } |
| 74 | + |
| 75 | + /// Return next not all false mask or request to read more data. |
| 76 | + fn next_mask(&mut self) -> VortexResult<Option<SplitMask>> { |
| 77 | + if !self.registered_splits.swap(true, Ordering::SeqCst) { |
| 78 | + let mut own_splits = BTreeSet::new(); |
| 79 | + self.reader.add_splits(0, &mut own_splits)?; |
| 80 | + self.static_splits.additional_splits(&mut own_splits)?; |
| 81 | + } |
| 82 | + |
| 83 | + // First consider masks we have previously started reading to return them in order |
| 84 | + while let Some(mask) = self.in_progress_masks.pop_front() { |
| 85 | + if let Some(read_mask) = self.read_mask(mask)? { |
| 86 | + return Ok(Some(read_mask)); |
| 87 | + } |
| 88 | + } |
| 89 | + |
| 90 | + // Lastly take next statically generated mask and perform read with it on our reader |
| 91 | + while let Some(mask) = self.static_splits.next() { |
| 92 | + match mask? { |
| 93 | + SplitMask::ReadMore(_) => { |
| 94 | + unreachable!("StaticSplitProducer never returns ReadMore") |
| 95 | + } |
| 96 | + SplitMask::Mask(m) => { |
| 97 | + if let Some(read_mask) = self.read_mask(m)? { |
| 98 | + return Ok(Some(read_mask)); |
| 99 | + } |
| 100 | + } |
| 101 | + } |
| 102 | + } |
| 103 | + Ok(None) |
| 104 | + } |
| 105 | +} |
| 106 | + |
| 107 | +impl MaskIterator for FilteringRowSplitIterator { |
| 108 | + fn additional_splits(&mut self, splits: &mut BTreeSet<usize>) -> VortexResult<()> { |
| 109 | + self.static_splits.additional_splits(splits) |
| 110 | + } |
| 111 | +} |
| 112 | + |
| 113 | +impl Iterator for FilteringRowSplitIterator { |
| 114 | + type Item = VortexResult<SplitMask>; |
| 115 | + |
| 116 | + fn next(&mut self) -> Option<Self::Item> { |
| 117 | + self.next_mask().transpose() |
| 118 | + } |
| 119 | +} |
| 120 | + |
| 121 | +pub struct FixedSplitIterator { |
| 122 | + splits: SplitState, |
| 123 | + row_mask: Option<RowMask>, |
| 124 | +} |
| 125 | + |
| 126 | +impl FixedSplitIterator { |
| 127 | + pub fn new(row_count: u64, row_mask: Option<RowMask>) -> Self { |
| 128 | + let mut splits = BTreeSet::new(); |
| 129 | + splits.insert(row_count as usize); |
| 130 | + Self { |
| 131 | + splits: SplitState::Splits(splits), |
| 132 | + row_mask, |
| 133 | + } |
| 134 | + } |
| 135 | +} |
| 136 | + |
| 137 | +impl MaskIterator for FixedSplitIterator { |
| 138 | + fn additional_splits(&mut self, splits: &mut BTreeSet<usize>) -> VortexResult<()> { |
| 139 | + match &mut self.splits { |
| 140 | + SplitState::Ranges(_) => { |
| 141 | + vortex_bail!("Can't insert additional splits if we started producing row ranges") |
| 142 | + } |
| 143 | + SplitState::Splits(s) => { |
| 144 | + s.append(splits); |
| 145 | + Ok(()) |
| 146 | + } |
| 147 | + } |
| 148 | + } |
| 149 | +} |
| 150 | + |
| 151 | +impl Iterator for FixedSplitIterator { |
| 152 | + type Item = VortexResult<SplitMask>; |
| 153 | + |
| 154 | + fn next(&mut self) -> Option<Self::Item> { |
| 155 | + match &mut self.splits { |
| 156 | + SplitState::Ranges(ranges) => { |
| 157 | + // Find next range that's not filtered out by supplied row_mask |
| 158 | + for (begin, end) in ranges { |
| 159 | + return if let Some(ref row_mask) = self.row_mask { |
| 160 | + if row_mask.slice(begin, end).is_empty() { |
| 161 | + continue; |
| 162 | + } |
| 163 | + Some(Ok(SplitMask::Mask(row_mask.slice(begin, end)))) |
| 164 | + } else { |
| 165 | + Some(Ok(SplitMask::Mask(RowMask::new_valid_between(begin, end)))) |
| 166 | + }; |
| 167 | + } |
| 168 | + None |
| 169 | + } |
| 170 | + SplitState::Splits(s) => { |
| 171 | + self.splits = SplitState::Ranges(Box::new( |
| 172 | + mem::take(s).into_iter().tuple_windows::<(usize, usize)>(), |
| 173 | + )); |
| 174 | + self.next() |
| 175 | + } |
| 176 | + } |
| 177 | + } |
| 178 | +} |
| 179 | + |
| 180 | +#[cfg(test)] |
| 181 | +mod tests { |
| 182 | + use std::collections::BTreeSet; |
| 183 | + |
| 184 | + use vortex_error::VortexResult; |
| 185 | + |
| 186 | + use crate::file::read::splits::{FixedSplitIterator, MaskIterator, SplitMask}; |
| 187 | + use crate::file::RowMask; |
| 188 | + |
| 189 | + #[test] |
| 190 | + #[should_panic] |
| 191 | + #[cfg_attr(miri, ignore)] |
| 192 | + fn register_after_start() { |
| 193 | + let mut mask_iter = FixedSplitIterator::new(10, None); |
| 194 | + mask_iter |
| 195 | + .additional_splits(&mut BTreeSet::from([0, 1, 2])) |
| 196 | + .unwrap(); |
| 197 | + assert!(mask_iter.next().is_some()); |
| 198 | + mask_iter |
| 199 | + .additional_splits(&mut BTreeSet::from([5])) |
| 200 | + .unwrap(); |
| 201 | + mask_iter.next(); |
| 202 | + } |
| 203 | + |
| 204 | + #[test] |
| 205 | + #[cfg_attr(miri, ignore)] |
| 206 | + fn filters_empty() { |
| 207 | + let mut mask_iter = |
| 208 | + FixedSplitIterator::new(10, Some(RowMask::try_new((4..6).collect(), 0, 10).unwrap())); |
| 209 | + mask_iter |
| 210 | + .additional_splits(&mut BTreeSet::from([2, 4, 6, 8, 10])) |
| 211 | + .unwrap(); |
| 212 | + assert_eq!( |
| 213 | + mask_iter |
| 214 | + .map(|split| split.map(|mask| match mask { |
| 215 | + SplitMask::ReadMore(_) => unreachable!("Will never read more"), |
| 216 | + SplitMask::Mask(m) => m, |
| 217 | + })) |
| 218 | + .collect::<VortexResult<Vec<_>>>() |
| 219 | + .unwrap(), |
| 220 | + vec![RowMask::new_valid_between(4, 6)] |
| 221 | + ); |
| 222 | + } |
| 223 | +} |
0 commit comments