scubatrace 0.9.4

Next-generation codebase analysis toolkit.

ScubaTrace

Next-Generation Codebase Analysis Toolkit.

Features

• Multi-Language Support (C, C++, Java, Python, JavaScript, Go)
• No Need To Compile
• Statement-Based AST Abstraction
• Code Call Graph
• Code Control Flow Graph
• Code Data/Control Dependency Graph
• References Inference
• CPG Based Multi-Granularity Slicing

Install

pip install scubatrace

Usage

Project-Level Analysis

Load a project (codebase)

proj = scubatrace.CProject("path/to/your/codebase", enable_lsp=True)

Call Graph

# Get the call graph of the project
callgraph = proj.callgraph
# Export call graph to a dot file
proj.export_callgraph("callgraph.dot")

Code Search

stat = proj.search_function("relative/path/to/your/file.c", start_line=20)

File-Level Analysis

Load a file from a project

file = proj.files["relative/path/to/your/file.c"]

Function-Level Analysis

Load a function from a file

the_first_func = file.functions[0]
func_in_tenth_line = file.function_by_line(10)

Call Relationships

def callers(self) -> dict[Function, list[Statement]]: ...
def callees(self) -> dict[Function, list[Statement]]: ...
def calls(self) -> list[Statement]: ...

Function Control Flow Graph

# Export the control flow graph to a dot file
func.export_cfg_dot("cfg.dot")

Function Data Dependency Graph

# Export the data dependency graph to a dot file
func.export_cfg_dot("ddg.dot", with_ddg=True)

Function Control Dependency Graph

# Export the control dependency graph to a dot file
func.export_cfg_dot("cdg.dot", with_cdg=True)

Function Code Walk

statements_you_interest = list(
    func.walk_backward(
        filter=lambda x: x.is_jump_statement,
        stop_by=lambda x: x.is_jump_statement,
        depth=-1,
        base="control",
    )
)
statements_you_interest = list(
    func.walk_forward(
        filter=lambda x: x.is_jump_statement,
        stop_by=lambda x: x.is_jump_statement,
        depth=-1,
        base="control",
    )
)

Multi-Granularity Slicing

# Slicing by lines
lines_you_interest = [4, 5, 19]
slice_statements = func.slice_by_lines(
    lines=lines_you_interest,
    control_depth=3,
    data_dependent_depth=5,
    control_dependent_depth=2,
)

# Slicing by statements
statements_you_interest = func.statements[0:3]
slice_statements = func.slice_by_statements(
    statements=statements_you_interest,
    control_depth=3,
    data_dependent_depth=5,
    control_dependent_depth=2,
)

Statement-Level Analysis

Load a statement from a function

the_first_stmt = the_first_func.statements[0]
stmt_in_second_line = the_first_func.statement_by_line(2)
stmt_by_type = func.statements_by_type('tree-sitter Queries', recursive=True)

Statement Controls

pre_controls: list[Statement] = stat.pre_controls
post_controls: list[Statement] = stat.post_controls

Statement Data Dependencies

pre_data_dependents: dict[Identifier, list[Statement]] = stat.pre_data_dependents
post_data_dependents: dict[Identifier, list[Statement]] = stat.post_data_dependents

Statement Control Dependencies

pre_control_dependents: list[Statement] = stat.pre_control_dependents
post_control_dependents: list[Statement] = stat.post_control_dependents

Statement References

references: dict[Identifier, list[Statement]] = stat.references

Statement Definitions

definitions: dict[Identifier, list[Statement]] = stat.definitions

Taint Analysis

# Check if the statement is tainted from function entry
is_taint_from_entry: bool = stat.is_taint_from_entry

AST Node

You can also get the AST node from a file, function, or statement.

file_ast = file.node
func_ast = func.node
stmt_ast = stat.node

ScubaTrace Landscape

Comparison with Other Tools

Tool	Type	Capabilities	Requires Compilation (Instruction)	Supported Languages	Limitations
ScubaTrace	Lib	CG/CFG/DataFlow/Slicing	✅ No	Multiple Languages
Soot	CLI/Lib (Java)	CG/CFG/DataFlow	❌ Yes	Java (Bytecode)	Cannot directly analyze the source code
LLVM	CLI/Lib (C)	CG/CFG/DataFlow	❌ Yes	C/C++ (IR)	Cannot directly analyze the source code
pycallgraph	CLI	CG	✅ No	Python	Does not provide a library, requires parsing the tool output
pycg	CLI	CG	✅ No	Python	Precision is low, requires parsing the tool output, no longer maintained
Jelly	CLI	CG	✅ No	JavaScript	Incomplete call graph (CG), the generated output requires further processing
Infer	OCaml	CG/CFG/DataFlow	❌ Yes	Multiple Languages	1. High cost of adaptation
CodeQL	QL	CG/CFG/DataFlow	❌ Required for compiled languages ✅ Not required for interpreted languages	Multiple Languages	1. Compiled languages require compilation 2. Requires learning QL and using it for analysis 3. Lower performance, slow for large-scale projects
Joern	CLI/Scala	CG/CFG/DataFlow	✅ No	Multiple Languages	1. The generated CG and other results cannot be directly used, require further processing 2. Generated CG graphs are prone to errors in resolving output failures 3. Lower performance, slow for large-scale projects