hohn/codeql
1
0
Fork 0
mirror of https://github.com/github/codeql.git synced 2026-04-26 17:25:19 +02:00
Code Issues Packages Projects Releases Wiki Activity

C++/Docs: expand on gvn and hashcons design

Browse Source
This commit is contained in:
Robert Marsh
2019-10-08 13:41:26 -07:00
parent aee87ebaaa
commit 07e7d061cd
1 changed files with 8 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
docs/language/learn-ql/cpp/value-numbering-hash-cons.rst
View File

@@ -2,7 +2,10 @@ Hash consing and value numbering
=================================================
Overview
--------
The C and C++ QL libraries include two different notions of expressions being the same. The hash consing library (defined in ``semmle.code.cpp.valuenumbering.HashCons``) provides a mechanism for identifying expressions that have the same syntactic structure. The value numbering library (defined in ``semmle.code.cpp.ir.ValueNumber``) provides a mechanism for identifying expressions that compute the same value at runtime.
In C and C++ QL databases, each node in the abstract syntax tree is represented by a separate object. This allows both analysis and results display to refer to specific appearances of a piece of syntax. However, it is frequently useful to determine whether two expressions are equivalent, either syntactically or semantically.
The hash consing library (defined in ``semmle.code.cpp.valuenumbering.HashCons``) provides a mechanism for identifying expressions that have the same syntactic structure. The value numbering library (defined in ``semmle.code.cpp.ir.ValueNumber``) provides a mechanism for identifying expressions that compute the same value at runtime.
Example C code
--------------
@@ -48,6 +51,10 @@ To get the ``GVN`` of an ``Expr``, use the ``globalValueNumber`` predicate.
While the ``GVN`` class has ``toString`` and ``getLocation`` methods, these are only provided as debugging aids. They give the ``toString`` and ``getLocation`` of an arbitrary ``Expr`` within the set.
Why not a predicate?
~~~~~~~~~~~~~~~~~~~~
The obvious interface for this library would be a predicate ``equivalent(Expr e1, Expr e2)``. However, this predicate would be very large, with a quadratic number of rows for each set of equivalent expressions. By using a class as an intermediate step, the number of rows can be kept linear, and therefore can be cached.
Example Queries
~~~~~~~~~~~~~~~