Compute space usage of a Searcher / SegmentReader / CompositeFile

src/common/composite_file.rs

@@ -4,6 +4,9 @@ use common::VInt;
 use directory::ReadOnlySource;
 use directory::WritePtr;
 use schema::Field;
+use space_usage::PerFieldSpaceUsage;
+use space_usage::ByteCount;
+use space_usage::FieldUsage;
 use std::collections::HashMap;
 use std::io::Write;
 use std::io::{self, Read};
@@ -166,6 +169,16 @@ impl CompositeFile {
             .get(&FileAddr { field, idx })
             .map(|&(from, to)| self.data.slice(from, to))
     }
+
+    pub fn space_usage(&self) -> PerFieldSpaceUsage {
+        let mut fields = HashMap::new();
+        for (&field_addr, &(start, end)) in self.offsets_index.iter() {
+            fields.entry(field_addr.field)
+                .or_insert_with(|| FieldUsage::empty(field_addr.field))
+                .add_field_idx(field_addr.idx, ByteCount(end - start));
+        }
+        PerFieldSpaceUsage::new(fields)
+    }
 }
 
 #[cfg(test)]

src/core/searcher.rs

@@ -5,6 +5,7 @@ use query::Query;
 use schema::Document;
 use schema::Schema;
 use schema::{Field, Term};
+use space_usage::SearcherSpaceUsage;
 use std::fmt;
 use std::sync::Arc;
 use termdict::TermMerger;
@@ -84,6 +85,15 @@ impl Searcher {
             .collect::<Vec<_>>();
         FieldSearcher::new(inv_index_readers)
     }
+
+    /// Summarize total space usage of this searcher.
+    pub fn space_usage(&self) -> SearcherSpaceUsage {
+        let mut space_usage = SearcherSpaceUsage::new();
+        for segment_reader in self.segment_readers.iter() {
+            space_usage.add_segment(segment_reader.space_usage());
+        }
+        space_usage
+    }
 }
 
 pub struct FieldSearcher {

src/core/segment_reader.rs

@@ -17,6 +17,8 @@ use schema::Document;
 use schema::Field;
 use schema::FieldType;
 use schema::Schema;
+use space_usage::SegmentSpaceUsage;
+use space_usage::ByteCount;
 use std::collections::HashMap;
 use std::fmt;
 use std::sync::Arc;
@@ -328,6 +330,20 @@ impl SegmentReader {
             .map(|delete_set| delete_set.is_deleted(doc))
             .unwrap_or(false)
     }
+
+    /// Summarize total space usage of this segment.
+    pub fn space_usage(&self) -> SegmentSpaceUsage {
+        SegmentSpaceUsage::new(
+            self.num_docs(),
+            self.termdict_composite.space_usage(),
+            self.postings_composite.space_usage(),
+            self.positions_composite.space_usage(),
+            self.fast_fields_composite.space_usage(),
+            self.fieldnorms_composite.space_usage(),
+            self.store_reader.space_usage(),
+            self.delete_bitset_opt.as_ref().map(|x| x.space_usage()).unwrap_or(ByteCount(0)),
+        )
+    }
 }
 
 impl fmt::Debug for SegmentReader {

src/fastfield/delete.rs

@@ -2,6 +2,7 @@ use bit_set::BitSet;
 use common::HasLen;
 use directory::ReadOnlySource;
 use directory::WritePtr;
+use space_usage::ByteCount;
 use std::io;
 use std::io::Write;
 use DocId;
@@ -62,6 +63,11 @@ impl DeleteBitSet {
             b & (1u8 << shift) != 0
         }
     }
+
+    /// Summarize total space usage of this bitset.
+    pub fn space_usage(&self) -> ByteCount {
+        ByteCount(self.data.len())
+    }
 }
 
 impl HasLen for DeleteBitSet {

src/lib.rs

@@ -195,6 +195,7 @@ pub(crate) mod fieldnorm;
 pub mod postings;
 pub mod query;
 pub mod schema;
+pub mod space_usage;
 pub mod store;
 pub mod termdict;
 

src/space_usage/mod.rs

@@ -0,0 +1,197 @@
+/*!
+Representations for the space usage of various parts of a Tantivy index.
+
+This can be used programatically, and will also be exposed in a human readable fashion in
+tantivy-cli.
+
+One important caveat for all of this functionality is that none of it currently takes storage-level
+details into consideration. For example, if your file system block size is 4096 bytes, we can
+under-count actual resultant space usage by up to 4095 bytes per file.
+*/
+
+use schema::Field;
+use std::collections::HashMap;
+use std::ops::{Add, AddAssign};
+
+/// Indicates space usage in bytes
+#[derive(Clone, Copy, Debug)]
+pub struct ByteCount(pub usize);
+
+impl Add for ByteCount {
+    type Output = ByteCount;
+    fn add(self, rhs: ByteCount) -> ByteCount {
+        ByteCount(self.0 + rhs.0)
+    }
+}
+
+impl AddAssign for ByteCount {
+    fn add_assign(&mut self, rhs: ByteCount) {
+        self.0 += rhs.0;
+    }
+}
+
+/// Represents combined space usage of an entire searcher and its component segments.
+#[derive(Clone, Debug)]
+pub struct SearcherSpaceUsage {
+    segments: Vec<SegmentSpaceUsage>,
+    total: ByteCount,
+}
+
+impl SearcherSpaceUsage {
+    pub(crate) fn new() -> SearcherSpaceUsage {
+        SearcherSpaceUsage {
+            segments: Vec::new(),
+            total: ByteCount(0),
+        }
+    }
+
+    /// Add a segment, to `self`.
+    /// Performs no deduplication or other intelligence.
+    pub(crate) fn add_segment(&mut self, segment: SegmentSpaceUsage) {
+        self.total += segment.total();
+        self.segments.push(segment);
+    }
+
+    /// Returns total byte usage of this searcher, including all large subcomponents.
+    /// Does not account for smaller things like `meta.json`.
+    pub fn total(&self) -> ByteCount {
+        self.total
+    }
+}
+
+/// Represents combined space usage for all of the large components comprising a segment.
+#[derive(Clone, Debug)]
+pub struct SegmentSpaceUsage {
+    num_docs: u32,
+
+    termdict: PerFieldSpaceUsage,
+    postings: PerFieldSpaceUsage,
+    positions: PerFieldSpaceUsage,
+    fast_fields: PerFieldSpaceUsage,
+    fieldnorms: PerFieldSpaceUsage,
+
+    store: StoreSpaceUsage,
+
+    deletes: ByteCount,
+
+    total: ByteCount,
+}
+
+impl SegmentSpaceUsage {
+    pub(crate) fn new(
+        num_docs: u32,
+        termdict: PerFieldSpaceUsage,
+        postings: PerFieldSpaceUsage,
+        positions: PerFieldSpaceUsage,
+        fast_fields: PerFieldSpaceUsage,
+        fieldnorms: PerFieldSpaceUsage,
+        store: StoreSpaceUsage,
+        deletes: ByteCount,
+    ) -> SegmentSpaceUsage {
+        let total = termdict.total()
+            + postings.total()
+            + positions.total()
+            + fast_fields.total()
+            + fieldnorms.total()
+            + store.total()
+            + deletes;
+        SegmentSpaceUsage {
+            num_docs,
+            termdict,
+            postings,
+            positions,
+            fast_fields,
+            fieldnorms,
+            store,
+            deletes,
+            total,
+        }
+    }
+
+    /// Total space usage in bytes for this segment.
+    pub fn total(&self) -> ByteCount {
+        self.total
+    }
+}
+
+/// Represents space usage for the Store for this segment.
+///
+/// This is composed of two parts.
+/// `data` represents the compressed data itself.
+/// `offsets` represents a lookup to find the start of a block
+#[derive(Clone, Debug)]
+pub struct StoreSpaceUsage {
+    data: ByteCount,
+    offsets: ByteCount,
+}
+
+impl StoreSpaceUsage {
+    pub(crate) fn new(data: ByteCount, offsets: ByteCount) -> StoreSpaceUsage {
+        StoreSpaceUsage { data, offsets }
+    }
+
+    /// Total space usage in bytes for this Store
+    pub fn total(&self) -> ByteCount {
+        self.data + self.offsets
+    }
+}
+
+/// Represents space usage for all of the (field, index) pairs that appear in a CompositeFile.
+///
+/// A field can appear with a single index (typically 0) or with multiple indexes.
+/// Multiple indexes are used to handle variable length things, where
+#[derive(Clone, Debug)]
+pub struct PerFieldSpaceUsage {
+    fields: HashMap<Field, FieldUsage>,
+    total: ByteCount
+}
+
+impl PerFieldSpaceUsage {
+    pub(crate) fn new(fields: HashMap<Field, FieldUsage>) -> PerFieldSpaceUsage {
+        let total = fields.values().map(|x| x.total()).fold(ByteCount(0), Add::add);
+        PerFieldSpaceUsage { fields, total }
+    }
+
+    /// Bytes used by the represented file
+    pub fn total(&self) -> ByteCount {
+        self.total
+    }
+}
+
+/// Represents space usage of a given field, breaking it down into the (field, index) pairs that
+/// comprise it.
+///
+/// See documentation for PerFieldSpaceUsage for slightly more information.
+#[derive(Clone, Debug)]
+pub struct FieldUsage {
+    field: Field,
+    weight: ByteCount,
+    /// A field can be composed of more than one piece.
+    /// These pieces are indexed by arbitrary numbers starting at zero.
+    /// `self.weight` includes all of `self.sub_weights`.
+    sub_weights: Vec<Option<ByteCount>>,
+}
+
+impl FieldUsage {
+    pub(crate) fn empty(field: Field) -> FieldUsage {
+        FieldUsage {
+            field,
+            weight: ByteCount(0),
+            sub_weights: Vec::new(),
+        }
+    }
+
+    pub(crate) fn add_field_idx(&mut self, idx: usize, size: ByteCount) {
+        if self.sub_weights.len() < idx {
+            self.sub_weights.resize(idx, None);
+        }
+        assert!(self.sub_weights[idx].is_none());
+        self.sub_weights[idx] = Some(size);
+        self.weight += size
+    }
+
+    /// Total bytes used for this field in this context
+    pub fn total(&self) -> ByteCount {
+        self.weight
+    }
+}

src/store/reader.rs

@@ -6,6 +6,8 @@ use datastruct::SkipList;
 use directory::ReadOnlySource;
 use lz4;
 use schema::Document;
+use space_usage::StoreSpaceUsage;
+use space_usage::ByteCount;
 use std::cell::RefCell;
 use std::io::{self, Read};
 use std::mem::size_of;
@@ -89,6 +91,11 @@ impl StoreReader {
         cursor = &cursor[..doc_length];
         Ok(Document::deserialize(&mut cursor)?)
     }
+
+    /// Summarize total space usage of this store reader.
+    pub fn space_usage(&self) -> StoreSpaceUsage {
+        StoreSpaceUsage::new(ByteCount(self.data.len()), ByteCount(self.offset_index_source.len()))
+    }
 }
 
 #[allow(needless_pass_by_value)]