Issue #2196 - EQL: delegating parsing to ANTLR #2209
Commits on Feb 23, 2024
-
-
-
#2196 Avoid duplication in generated ANTLR grammar due to multiple pa…
…rameterisations of the same terminal Since we don't include the parameterisation of terminals in the ANTLR grammar we need to eliminate duplicate tokens.
-
-
-
-
-
Commits on Feb 27, 2024
Commits on Feb 28, 2024
-
#2196 Configure ANTLR Maven plugin to output artifacts into platform-…
…pojo-bl Also, don't generate a listener, which is not needed.
-
-
-
#2196 Initial transition to ANTLR tokens in fluent API implementation
- Replaced the intermediate Tokens class by EqlSentenceBuilder. - Temporarily commented out the old code which won't compile anymore and is staged for removal.
-
#2196 Move the instantiation of ITypeCasting associated with "end" of…
… "case when" to a later stage Instead of doing it in the fluent API implementation, we will do it during the compilation of expressions.
-
-
-
#2196 Adjust ANTLR grammar to correctly handle yieldAll cases
1. Allow zero repetitions of `yield` following `yieldAll` 2. Give precedence to the `model` rule over `anyYield`. This will correctly apply rule `model` when there are no yields.
Commits on Feb 29, 2024
-
-
#2196 Collect tokens for terminal methods in the fluent API
E.g., `model()`, `modelAsEntitty(...)` They need to be collected since they are part of the grammar. Without them EQL sentences would be rejected by the parser.
-
#2196 Replace List<Token> with a higher level abstraction to represen…
…t sources of ANTLR tokens
-
#2196 Collect starting tokens for standalone condition/expression and…
… order by For the same reason that we collect terminal tokens (e.g., `model`)
-
Commits on Mar 1, 2024
-
#2196 Simplify the
selectgrammar ruleOn the grammar level all `select`s are equal as there is no information about token parameters, thus there is no need to distinugish sourceless selects (which were also introducing ambiguity and slowing down the parser).
-
#2196 Send whitespace and comments to a hidden channel instead of ski…
…pping them This makes error messages more readable since they will include these tokens
-
-
-
-
-
#2196 Move common token-handling code to the abstract ANTLR visitor
This is a part of migration from AbstractTokenBuilder.
-
#2196 Enhance stage 1 Yield structures with static factory methods an…
…d additional constructors
-
-
#2196 Implement compilation of some single operands
Value, Param, Prop, Round, Expr
-
-
-
-
-
-
-
-
#2196 Migrate users of token builders for standalone expression/condi…
…tion to the new API for compiling EQL
-
Commits on Mar 4, 2024
-
-
-
-
-
-
#2196 Implement compilation of various query sources
This includes: * entity type (persisted/synthetic/union) * query models
-
-
-
#2196 #2178 Extend the yield operand rule with expressions involving …
…yield operands This allows the body of expressions to use yield operands, not just single operands.
-
-
-
#2196 Implement compilation of comparison operators in a generic way
ComparisonOperandVisitor allows to handle both single and multi-operands uniformly.
Commits on Mar 5, 2024
-
#2196 Be consistent with source ID generation logic when compiling po…
…tential query models as operands
-
-
-
-
#2196 Avoid ambiguity in the parser related to yielded expressions
Yield operands can be said to extend regular single operands. This gives rise to an interesting case: yielded expressions (delimited by `beginExpr` & `endExpr`) can contain the whole set of yield operands, not just single operands. Therefore, we need a separate rule for yielded expressions. However, if it used the same pair of delimiters (i.e., started with `beginExpr`), an ambiguity would occur: is this a single operand or a yield operand? This could be resolved with ANTLR's semantic predicates, but introducing a new pair of delimiters (the chosen approach) is much simpler. N.B.: ENDYIELDEXPR was introduced for the sake of consistency, we could have used BEGINYIELDEXPR ... ENDEXPR as well.
-
-
-
-
-
-
-
-
Commits on Mar 6, 2024
-
-
#2196 Adjust fluent API implementation to collect the correct token w…
…hen ending expressions inside yields
-
-
Commits on Mar 7, 2024
Commits on Mar 8, 2024
-
-
-
Introduce a Collection utility that allows to build any collection in…
… a single method call chain
-
-
#2196 Splice AND/OR conditions during compilation
To align with EQL's stage 1 structures for AND/OR conditions we need to transform ANTLR's parse results which are in the form of a binary tree. Example: (c1 OR (c2 OR c3)) -> (c1 OR c2 OR c3) This transformation does not affect the semantics, but it reduces the depth of resulting trees.
-
#2196 Modify test assertions that had previously relied on a particul…
…ar structure of query conditions
-
Commits on Mar 13, 2024
-
-
#2196 Disable maven-enforce-plugin for platform-benchmark module
It is not applicable here
-
-
#2196 ANTLR grammar: optimise the rules associated with yields
Yield1 and YieldMany were merged into a single alternation to avoid long lookaheads, which had been observed to be as much as 73 tokens. Long lookaheads were caused by the need to distinguish between a *single yield* or a *first aliased yield*, which required the "as" token to be encountered before a decision could be made.
Commits on Mar 14, 2024
-
-
-
#2196 Move the executable
mainmethod from CanonicalEqlGrammar into…… a separate class Keeps things decoupled
-
-
Commits on Mar 15, 2024
-
-
-
#2196 Cleanup: use util methods for brevity and add documentation
No semantics have been changed
-
#2196 Adjust failing tests to rely on the actual order of source ID g…
…eneration Ideally these tests should not rely on implementation details such as ID generation or should be alltogether replaced by equi-testing.
-
#2196 Avoid NPE when copying a List that might contain nulls
Ideally we should replace all occurences of nulls with some unique Object instance or a custom Null type singleton so that we never have to deal with "null as Value" during query transformation.
-
#2196 Add a test to cover usage of nulls in
{any,all}OfValues(...)……and `values(...)`
-
-
#2196 Limit the visibility of
Value1constructor and use factory me……thods instead
-
Commits on Mar 18, 2024
-
-
#2196 Implement
equals()of query models in terms of tokens provide……d by the underlying token source
-
-
-
Commits on Mar 22, 2024
Commits on Mar 24, 2024
Commits on Mar 26, 2024
Commits on Mar 28, 2024
Commits on Apr 1, 2024
Commits on Apr 2, 2024
Commits on Apr 10, 2024
Commits on Apr 11, 2024
-
#2196 Prioritise direct ANTLRv4 dependency over the transitive one
The transitive dependency has a different version, coming from the GraphQL dependency. In TG applications the transitive was being picked up instead of the direct one.
Commits on Apr 12, 2024
Commits on Apr 26, 2024
Commits on May 21, 2024
Commits on Jun 22, 2024
Commits on Jun 28, 2024
Commits on Aug 7, 2024
Commits on Aug 8, 2024
-
#2196 Javadoc adjustments, EQL grammar DLS improvements.
Reduced type visiblity, adjusted type parameterisation, method rationalisation.
-
-
#2196 #2178 EQL BNF grammar: swap startDate & endDate labels in DateD…
…iffInterval Follow-up on 5227200
-
#2196 #2178 EQL BNF grammar: remove redundant labels
Follow-up on dc70454
-
#2196 #2178 EQL BNF grammar: optimise yield rules
Follow-up on e1a88c6
-
#2196 #2178 EQL BNF grammar: remove labels that cause errors in ANTLR…
… grammar Can't apply the same label to elements of different types.
-
#2196 #2178 Introduce interface GrammarTransformer
This will make it easier to compose transformations.
-
#2196 #2178 EQL BNF grammar: align comparisonOperand rule with ANTLR …
…grammar Follow-up on 4433838
-
#2196 #2178 Avoid redundant wrapping of terms in Sequence
1. Hide constructors of Sequence and use efficient static methods. 2. Modify the structure of Alternation to use a list of Term instead of a list of Sequence.
-
-
-
Commits on Aug 9, 2024
-
#2196 #2178 Rename TermMetadata to Metadata
Generalise the concept of metadata, which can be applied to any part of BNF, not limited to terms.
-
-
-
#2196 #2178 Refactor metadata representation from key-value pairs to …
…annotation types
-
-
-
#2196 Align SelectEnd rule with the ANTLR grammar
Follow-up on b7e5d7d
-
#2196 Don't inline rules by default, use a special annotation
Inlining by default goes wrong when we begin inlining more aggressively, which is required for the SelectSource rule: SelectFrom = SelectSource as? Join? Where? GroupBy? SelectEnd SelectSource = select We would like to inline SelectSource here, but it was not considered inlineable as it appears in a sequence of other terms. This would require us to perform aggressive inlining, which could have too far-reaching effects (e.g., also inlining Join and Where), which could be addressed with more "NoInline" annotations, but it would get tedious. With this change we reverse direction w.r.t. inlining - don't do it by default, but only for rules that are explicitly asking for it. An upcoming change will add support for aggressive inlining.
-
-
#2196 Add a transformation for the Select rule
Follow-up on 18312d5
-
#2196 Try to maintain the order of rules from the canonical grammar i…
…n the generated ANTLR grammar
-
#2196 Add a transformation for the YieldOperandExpr rule to align wit…
…h the ANTLR grammar Follow-up on 9e03459
Commits on Aug 12, 2024
-
-
-
#2196 BNF-to-ANTLR: replace string manipulation for token labels by a…
… grammar transformer Also remove the heuristic for alternative labels. The corresponding annotation should be used instead.
-
-
#2196 Label the select token in the SelectSource rule to align with t…
…he ANTLR grammar
-
-
-
-
Commits on Aug 13, 2024
Commits on Aug 14, 2024
Commits on Aug 15, 2024
-
-
-
-
-
-
#2196 Adjusted an odd, rundomly failing test, to use the exact same e…
…xpression for both expected and actual values.
-
#2196 Adjusted an odd, rundomly failing test, to use the exact same e…
…xpression for both expected and actual values.
-
-
-