Merge #200

200: Implement negation for non-existing nodes r=thomaskrause a=thomaskrause

This implements #187 but does not implement the more complex structures (as described in #199) yet. Also note that we changed form a prefix ? to a suffix, to make it easier to implement optional parts of the query term in the AQL parser later on when needed.

We implement these negated operators with non-existent operands as unary operators (with the existing node as the one that is filtered using the non-existent expression). There is also some refactoring and renaming to make it easier to use the value filter functions outside the conjunction code (in our operator) to avoid duplicated implementation of the node search logic.

Co-authored-by: Thomas Krause <[email protected]>
Co-authored-by: Thomas Krause <[email protected]>

CHANGELOG.md

@@ -5,11 +5,16 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Added
+
+- Added generic operator negation without existence assumption, 
+  if only one side of the negated operator is optional  (#187).
+
 ## [1.1.0] - 2021-09-09
 
 ### Added
 
-- Added generic operator negation with existance assumption by adding `!` before the binary operator (#186)
+- Added generic operator negation with existence assumption by adding `!` before the binary operator (#186)
 
 ### Changed
 

core/src/annostorage/inmemory.rs

@@ -1,4 +1,4 @@
-use super::{AnnotationStorage, Match, MatchGroup};
+use super::{AnnotationStorage, Match};
 use crate::annostorage::ValueSearch;
 use crate::errors::Result;
 use crate::malloc_size_of::MallocSizeOf;
@@ -8,6 +8,7 @@ use crate::{annostorage::symboltable::SymbolTable, errors::GraphAnnisCoreError};
 use core::ops::Bound::*;
 use itertools::Itertools;
 use rustc_hash::{FxHashMap, FxHashSet};
+use smallvec::SmallVec;
 use smartstring::alias::String;
 use std::borrow::Cow;
 use std::collections::{BTreeMap, HashMap};
@@ -407,11 +408,11 @@ where
         ns: Option<&str>,
         name: Option<&str>,
         it: Box<dyn Iterator<Item = T>>,
-    ) -> MatchGroup {
+    ) -> SmallVec<[Match; 8]> {
         if let Some(name) = name {
             if let Some(ns) = ns {
                 // return the only possible annotation for each node
-                let mut matches = MatchGroup::new();
+                let mut matches = SmallVec::<[Match; 8]>::new();
                 let key = Arc::from(AnnoKey {
                     ns: ns.into(),
                     name: name.into(),
@@ -441,7 +442,7 @@ where
                     })
                     .collect();
                 // return all annotations with the correct name for each node
-                let mut matches = MatchGroup::new();
+                let mut matches = SmallVec::<[Match; 8]>::new();
                 for item in it {
                     for (key_symbol, key) in matching_key_symbols.iter() {
                         if let Some(all_annos) = self.by_container.get(&item) {
@@ -458,7 +459,7 @@ where
             }
         } else {
             // return all annotations for each node
-            let mut matches = MatchGroup::new();
+            let mut matches = SmallVec::<[Match; 8]>::new();
             for item in it {
                 let all_keys = self.get_all_keys_for_item(&item, None, None);
                 for anno_key in all_keys {

core/src/annostorage/mod.rs

@@ -155,13 +155,13 @@ where
 
     /// Get the matching annotation keys for each item in the iterator.
     ///
-    /// This function allows to filter the received annotation keys by the specifying the namespace and name.
+    /// This function allows to filter the received annotation keys by specifying the namespace and name.
     fn get_keys_for_iterator(
         &self,
         ns: Option<&str>,
         name: Option<&str>,
         it: Box<dyn Iterator<Item = T>>,
-    ) -> MatchGroup;
+    ) -> SmallVec<[Match; 8]>;
 
     /// Return the total number of annotations contained in this `AnnotationStorage`.
     fn number_of_annotations(&self) -> usize;

core/src/annostorage/ondisk.rs

@@ -8,15 +8,14 @@ use crate::util::disk_collections::{DiskMap, EvictionStrategy};
 use crate::util::{self, memory_estimation};
 use core::ops::Bound::*;
 use rand::seq::IteratorRandom;
+use smallvec::SmallVec;
 use std::borrow::Cow;
 use std::collections::{BTreeMap, HashMap, HashSet};
 use std::path::{Path, PathBuf};
 use std::sync::Arc;
 
 use smartstring::alias::String as SmartString;
 
-use super::MatchGroup;
-
 pub const SUBFOLDER_NAME: &str = "nodes_diskmap_v1";
 
 const UTF_8_MSG: &str = "String must be valid UTF-8 but was corrupted";
@@ -464,15 +463,15 @@ where
         ns: Option<&str>,
         name: Option<&str>,
         it: Box<dyn Iterator<Item = T>>,
-    ) -> MatchGroup {
+    ) -> SmallVec<[Match; 8]> {
         if let Some(name) = name {
             if let Some(ns) = ns {
                 // return the only possible annotation for each node
                 let key = Arc::from(AnnoKey {
                     ns: ns.into(),
                     name: name.into(),
                 });
-                let mut matches = MatchGroup::new();
+                let mut matches = SmallVec::<[Match; 8]>::new();
                 if let Some(symbol_id) = self.anno_key_symbols.get_symbol(&key) {
                     // create a template key
                     let mut container_key = create_by_container_key(T::default(), symbol_id);
@@ -500,7 +499,7 @@ where
                     })
                     .collect();
                 // return all annotations with the correct name for each node
-                let mut matches = MatchGroup::new();
+                let mut matches = SmallVec::<[Match; 8]>::new();
                 for item in it {
                     for (container_key, anno_key) in matching_qnames.iter_mut() {
                         // Set the first bytes to the ID of the item.

graphannis/src/annis/db/aql/ast.rs

@@ -37,6 +37,7 @@ pub enum Literal {
         spec: NodeSearchSpec,
         pos: Option<Pos>,
         variable: Option<String>,
+        optional: bool,
     },
     BinaryOp {
         lhs: Operand,
@@ -63,6 +64,7 @@ pub enum Operand {
         spec: Rc<NodeSearchSpec>,
         pos: Pos,
         variable: Option<String>,
+        optional: bool,
     },
 }