Status: The toolset is still undergoing a major rewrite. Consider this toolkit as "pre-alpha". Old tools are being removed, and new ones are being added. Features are being added, while bugs are constantly being fixed. The XPath/XQuery engine is still being rewritten.
The repo g4-scripts contains a collections of Bash which use Trash. The repo also contains XQuery scripts that implement complex operations on a parse tree. You can also read about Trash details in my blog.
Trash is a collection of ~40 command-line tools to analyze and transform Antlr parse trees and grammars. The toolkit can: generate a parser application for an Antlr4 grammar for any target and any OS; analyze the grammar for common problems; automate changes applied to a grammar scraped from a specification; transform parse trees for transpilating and proprocessing source code. With the Antlr toolkit and the collection of Antlr grammars, one can write programming language tools quickly and easily.
The toolkit is designed around a JSON representation of parse trees and command-line tools that read, modify, and write those tree via standard input and output. Complex refactorings can be achieved by chaining different commands together.
Each app in Trash
is implemented as a Dotnet Tool console application, and can be used on Windows, Linux, or Mac.
No prerequisites are required other than installing the
NET SDK, and the toolchains
for any other targets you want to use.
The toolkit uses Antlr and XPath2. The code is implemented in C#.
An application of the toolkit was used to scrape and refactor the Dart2 grammar from spec. See this script.
Copy this script and execute it in a command-line prompt.
dotnet tool install -g trcaret
dotnet tool install -g trclonereplace
dotnet tool install -g trcombine
dotnet tool install -g trconvert
dotnet tool install -g trcover
dotnet tool install -g trfoldlit
dotnet tool install -g trgen
dotnet tool install -g trgenvsc
dotnet tool install -g trglob
dotnet tool install -g triconv
dotnet tool install -g tritext
dotnet tool install -g trjson
dotnet tool install -g trparse
dotnet tool install -g trperf
dotnet tool install -g trquery
dotnet tool install -g trrename
dotnet tool install -g trsort
dotnet tool install -g trsplit
dotnet tool install -g trsponge
dotnet tool install -g trtext
dotnet tool install -g trtokens
dotnet tool install -g trtree
dotnet tool install -g trunfold
dotnet tool install -g trwdog
dotnet tool install -g trxml
dotnet tool install -g trxml2
dotnet tool uninstall -g trcaret
dotnet tool uninstall -g trclonereplace
dotnet tool uninstall -g trcombine
dotnet tool uninstall -g trconvert
dotnet tool uninstall -g trcover
dotnet tool uninstall -g trfoldlit
dotnet tool uninstall -g trgen
dotnet tool uninstall -g trgenvsc
dotnet tool uninstall -g trglob
dotnet tool uninstall -g triconv
dotnet tool uninstall -g tritext
dotnet tool uninstall -g trjson
dotnet tool uninstall -g trparse
dotnet tool uninstall -g trperf
dotnet tool uninstall -g trrename
dotnet tool uninstall -g trsort
dotnet tool uninstall -g trsplit
dotnet tool uninstall -g trsponge
dotnet tool uninstall -g trtext
dotnet tool uninstall -g trtokens
dotnet tool uninstall -g trtree
dotnet tool uninstall -g trunfold
dotnet tool uninstall -g trwdog
dotnet tool uninstall -g trxml
dotnet tool uninstall -g trxml2
NB: Out of date
- tranalyze -- Analyze a grammar
- trcombine -- Combine a split Antlr4 grammar
- trconvert -- Convert a grammar from one for to another
- trdot -- Print a parse tree in Graphvis Dot format
- trenum -- Not functional, to enumerate strings from grammar.
- trfirst -- Outputs first sets of a grammar
- trfold -- Perform fold transform on a grammar
- trfoldlit -- Perform fold transform on grammar with literals
- trformat -- Format a grammar
- trgen -- Generate an Antlr4 parser for a given target language
- trgen2 -- Generate files from template and XML doc list.
- trgroup -- Perform a group transform on a grammar
- tritext -- Get strings from a PDF file
- trjson -- Print a parse tree in JSON structured format
- trkleene -- Perform a Kleene transform of a grammar
- trmove -- Move nodes in a parse tree
- trparse -- Parse a grammar or use generated parse to parse input
- trperf -- Perform performance analysis of an Antlr grammar parse
- trpiggy -- Perform a parse tree rewrite
- trprint -- Print a parse tree, including off-token characters
- trrename -- Rename symbols in a grammar
- trrr -- (No description.)
- trrup -- Remove useless parentheses in a grammar
- trsem -- Read static semantics and generate code
- trsort -- Sort rules in a grammar
- trsplit -- Split a combined Antlr4 grammar
- trsponge -- Extract parsing results output of Trash command into files
- trst -- Print a parse tree in Antlr4 ToStringTree()
- trstrip -- Strip a grammar of all actions, labels, etc.
- trtext -- Print a parse tree with a specific interval
- trthompson -- (No description.)
- trtokens -- Print tokens in a parse tree
- trtree -- Print a parse tree in a human-readable format
- trull -- Transform a grammar with upper- and lowercase string literals
- trunfold -- Perform an unfold transform on a grammar
- trungroup -- Perform an ungroup transform on a grammar
- trwdog -- Kill a program that runs too long
- trxml -- Print a parse tree in XML structured format
- trxml2 -- Print an enumeration of all paths in a parse tree to leaves
git clone https://github.com/antlr/grammars-v4
cd grammars-v4/python/python
trparse *.g4 | trquery 'grep //grammarDecl' | trtext
# Output:
# PythonLexer.g4:lexer grammar PythonLexer;
# PythonParser.g4:parser grammar PythonParser;
trgen
cd Generated
dotnet build
cat - <<EOF | trparse | trquery 'grep //test' | trtext
x == y
x == y if z == b else a == u
lambda: a
lambda x, y: a
EOF
# Output:
# a
# lambda x, y: a
# a
# lambda: a
# a == u
# x == y if z == b else a == u
# x == y
trparse -i "a == b" | trtree
trtree
is only one of several ways to view parse tree data.
Other programs for different output are
trjson for JSON output,
trxml for XML output,
trst for Antlr runtime ToStringTree output,
trdot,
trprint for input text for the parse,
and
tragl.
trparse ada.g2 | trconvert | trprint | less
This command parses an old Antlr2 grammar using trparse, converts the parse tree data to Antlr4 syntax using trconvert and finally prints out the converted parse tree data, ada.g4 using trprint. Other grammar that can be converted are Antlr3, Bison, and ISO EBNF. In order to use the grammar to parse data, you will need to convert it to an Antlr4 grammar.
mkdir foobar; cd foobar; trgen
This command creates a parser application for the C# target.
If executed in an empty directory, which is done in the example
shown above, trgen
creates an application using the Arithmetic grammar.
If executed in a directory containing
a Antlr Maven plugin (pom.xml
), trgen
will create a program according
to the information specified in the pom.xml
file. Either way, it creates a directory
Generated/
, and places the source code there.
trgen
has many options to generate a parser from any Antlr4 grammar, for any target.
But, if a parser is generated for the C# target, built using the NET SDK, then trparse
can execute the generated parser, and can be used with all the other tools in Trash. _NB:
In order to use the generate parser application, you must first build it:
dotnet restore Generated/Test.csproj
dotnet build Generated/Test.csproj
trparse -i "1+2+3" | trtree
After using trgen
to generate a parser program in C#, shown previously,
and after building the program, you can run the parser using trparse
. This program
looks for the generated parser in directory Generated/
. If it exists,
it will run the parser application in the directory. You can pass
as command-line arguments an input string or input file. If no command-line
arguments are supplied, the program will read stdin. The output of trparse
, as
with most tools of Trash, is parse tree data.
mkdir empty; cd empty; trgen; dotnet build Generated/Test.csproj; \
trparse -i "1+2+3" | trquery "grep //SCIENTIFIC_NUMBER" | trst
With this command, a directory is created, the Arithmetic grammar generated, build,
and then run using trparse.
The trparse
tool unifies all parsing, whether it's parsing a grammar or parsing input
using a generated parser application. The output from the trparse
tool is a parse
tree which you can search. Trquery
is the generalized search program for parse trees. Trquery
uses XPath expressions to
precisely identify nodes in the parse tree.
XPath was added to Antlr4, but Trash
takes the idea
further with the addition of an XPath2 engine ported from the
Eclipse Web toolkit.
XPath is a well-defined language that should be
used more often in compiler construction.
trparse Arithmetic.g4 | trrename "//parserRuleSpec//labeledAlt//RULE_REF[text() = 'expression']" "xxx" | trtext > new-source.g4
trparse Arithmetic.g4 | trrename -r "expression,expression_;atom,atom_;scientific,scientific_" | trprint
In these two examples, the Arithmetic grammar is parsed.
trrename reads the parse tree data and
modifies it by renaming the expression
symbol two ways: first by XPath expression identifying the LHS terminal
symbol of the expression
symbol, and the second by assumption that the tree is an Antlr4 parse tree,
then renaming a semi-colon-separated list of paired renames. The resulting code is reconstructed and saved.
trrename
does not rename symbols in actions, nor does it rename identifiers corresponding to the
grammar symbols in any support source code (but it could if the tool is extended).
git clone https://github.com/antlr/grammars-v4.git; \
cd grammars-v4/java/java9; \
trgen; dotnet build Generated/Test.csproj;\
trparse examples/AllInOne8.java | trquery "greap //methodDeclaration" | trst | wc
This command clones the Antlr4 grammars-v4 repo, generates a parser for the Java9 grammar,
then runs the parser on examples/AllInOne8.java.
The parse tree is then piped to trquery
to find all parse tree nodes that are
a methodDeclaration
type, converts it to a simple string, and counts the result using
wc
.
trparse Java9.g4 | trstrip | trtext > Essential-Java9.g4
Since Antlr2, one can written a combined parser/lexer in one file, or a split parser/lexer in two files. While it's not hard to split or combine a grammar, it's tedious. For automating transformations, it's necessary because Antlr4 requires the grammars to be split when super classes are needed for different targets.
trcombine ArithmeticLexer.g4 ArithmeticParser.g4 | trprint > Arithmetic.g4
This command calls trcombine which parses two split grammar files ArithmeticLexer.g4 and ArithmeticParser.g4, and creates a combined grammar for the two.
trparse Arithmetic.g4 | trsplit | trsponge -o true
This command calls trsplit which splits the grammar into two parse tree results, one that defines ArithmeticLexer.g4 and the other that defines ArithmeticParser.g4. The tool trsponge is similar to the tee in Linux: the parse tree data is split and placed in files.
A parsing result set is a JSON serialization of an array of:
- A set of parse tree nodes.
- Parser information related to the parse tree nodes.
- Lexer information related to the parse tree nodes.
- The name of the input corresponding to the parse tree nodes.
- The input text corresponding to the parse tree nodes.
Most commands in Trash read and/or write parsing result sets.
Grammars | File suffix |
---|---|
Antlr4 | .g4 |
Antlr3 | .g3 |
Antlr2 | .g2 |
Bison | .y |
LBNF | .cf |
W3C EBNF | .ebnf |
ISO 14977 | .iso14977, .iso |
Trash provides a number of transformations that can help to make grammars cleaner (reformatting), more readable (reducing the length of the RHS of a rule), and more efficient (reducing the number of non-terminals) for Antlr.
Some of these refactorings are very specific for Antlr due to the way the parser works, e.g., converting a prioritized chain of productions recognizing an arithmetic expression to a recursive alternate form. The refactorings implemented are:
- Remove useless parentheses
- Remove useless parser rules
- Rename lexer or parser symbol
- Unfold
- Group alts
- Ungroup alts
- Upper and lower case string literals
- Fold
- Replace direct left recursion with right recursion
- Replace direct left/right recursion with Kleene operator
- Replace indirect left recursion with right recursion
- Replace parser rule symbols that conflict with Antlr keywords
- Replace string literals in parser with lexer symbols
- Replace string literals in parser with lexer symbols, with lexer rule create
- Delabel removes the annoying and mostly useless labeling in an Antlr grammar
The source code for the extension is open source, free of charge, and free of ads. For the latest developments on the extension, check out my blog.
git clone https://github.com/kaby76/Trash
cd Trash
make clean; make; make install
You must have the NET SDK version 8 installed to build and run.
See https://github.com/kaby76/Trash/releases.
If you have any questions, email me at ken.domino gmail.com