Skip to content

Rewrite Code

One of the powers of ast-grep is that it can not only find code patterns, but also transform them into new code.

For example, you may want to rename a variable, change a function call, or add a comment. ast-grep provides two ways to do this: using the --rewrite flag or using the fix key in YAML rules.

Using sg run -p 'pat' --rewrite

The simplest way to rewrite code is to use the --rewrite flag with the sg run command. This flag takes a string argument that specifies the new code to replace the matched pattern. For example, if you want to change all occurrences of identifier foo to bar, you can run:

sg run --pattern 'foo' --rewrite 'bar' --lang python

This will show you a diff of the changes that will be made. If you are using interactive mode by the --interactive flag, ast-grep ask you if you want to apply them.


You can also use the --update-all or -U flag to automatically accept the changes without confirmation.

Using fix in YAML Rule

Another way to rewrite code is to use the fix option in a YAML rule file. This option allows you to specify more complex and flexible rewrite rules, such as using transformations and regular expressions.

Let's look at a simple example of using fix in a YAML rule (playground Link). Suppose we have a Python file named with the following content:

def foo(x):
    return x + 1

y = foo(2)

We want to only change the name of the function foo to baz, but not variable/method/class. We can write a YAML rule file named change_func.yml with the following content:

id: change_def
language: Python
  pattern: |
    def foo($X):
fix: |-
  def baz($X):

--- # this is YAML doc separator to have multiple rules in one file

id: change_param
  pattern: foo($X)
fix: baz($X)

The first rule matches the definition of the function foo, and replaces it with baz. The second rule matches the calls of the function foo, and replaces them with baz. Note that we use $X and $$$S as meta variables, which can match any expression and any statement, respectively. We can run:

sg scan -r change_func.yml

This will show us the following diff:

def foo(x): 
def baz(n): 
    return n + 1

y = foo(2) 
y = baz(2) 

We can see that the function name and parameter name are changed as we expected.

Pro Tip

You can have multiple rules in one YAML file by using the YAML document separator ---. This allows you to group related rules together!

Use Meta Variable in Rewrite

As we saw in the previous example, we can use meta variables in both the pattern and the fix parts of a YAML rule. They are like regular expression capture groups. Meta variables are identifiers that start with $, and they can match any syntactic element in the code, such as expressions, statements, types, etc. When we use a meta variable in the fix part of a rule, it will be replaced by whatever it matched in the pattern part.


non-matched meta-variable will be replaced by empty string in the fix.

For example, if we have a rule like this:

id: swap
language: Python
  pattern: $X = $Y
fix: $Y = $X

This rule will swap the left-hand side and right-hand side of any assignment statement. For example, if we have a code like this:

a = b
c = d + e
f = g * h

The rule will rewrite it as:

b = a
d + e = c
g * h = f

Playground link

Note that this may not be a valid or sensible code transformation, but it illustrates how meta variables work.

Rewrite is Indentation Sensitive

ast-grep's rewrite is indentation sensitive. That is, the indentation level of a meta-variable in the fix string is preserved in the rewritten code.

For example, if we have a rule like this:

id: lambda-to-def
language: Python
  pattern: '$B = lambda: $R'
fix: |-
  def $B():
    return $R

This rule will convert a lambda function to a standard def function. For example, if we have a code like this:

b = lambda: 123

The rule will rewrite it as:

def b():
  return 123

Note that the indentation level of return $R is preserved as two spaces in the rewritten code, even if the replacement 123 in the original code does not have indentation at all.

fix string's indentation is preserved relative to their position in the source code. For example, if the lambda appears within if statement, the diff will be like:

if True:
    c = lambda: 456
    def c():     
      return 456

Note that the return 456 line has an indentation of four spaces. This is because it has two spaces indentation as a part of the fix string, and two additional spaces because the fix string as a whole is inside the if statement in the original code.

Use transform in Rewrite

Sometimes, we may want to apply some transformations to the meta variables in the fix part of a YAML rule. For example, we may want to change the case, add or remove prefixes or suffixes. ast-grep provides a transform key that allows us to specify such transformations.

transform accepts a dictionary of which:

  • the key is the new variable name to be introduced and
  • the value is a transformation object that specifies which meta-variable is transformed and how.

A transformation object has a key indicating which string operation will be performed on the meta variable, and the value of that key is another object (usually with the source key). Different string operation keys expect different object values.

Converting generator expression to list comprehension in Python is a good example to illustrate transform.

More concretely, we want to achieve diffs like below:

"".join(i for i in iterable) 
"".join([i for i in iterable]) 

This rule will convert the generator inside join to a list.

id: convert_generator
  kind: generator_expression
  pattern: $GEN
transform:            # 1. the transform option
  LIST:               # 2. New variable name
    substring:        # 3. the transform operation name
      source: $GEN    # 4.1 transformation source
      startChar: 1    # 4.2 transformation argument
      endChar: -1
fix: '([$LIST])'      # 5. use the new variable in fix

Let's discuss the API step by step:

  1. The transform key is used to define one or more transformations that we want to apply to the meta variables in the pattern part of the rule.
  2. The LIST key is the new variable name that we can use in fix or later transformation. We can choose any name as long as it does not conflict with any existing meta variable names. Note, the new variable name does not start with $.
  3. The substring key is the transform operation name that we want to use. This operation will extract a substring from the source string based on the given start and end characters.
  4. substring accepts an object
    1. The source key specifies which meta variable we want to transform. It should have $ prefix. In this case, it is $GEN that which matches the generator expression in the code.
    2. The startChar and endChar keys specify the indices of the start and end characters of the substring that we want to extract. In this case, we want to extract everything except the wrapping parentheses, which are the first and last characters: ( and ).
  5. The fix key specifies the new code that we want to replace the matched pattern with. We use the new variable $LIST in the fix part, and wrap it with [ and ] to make it a list comprehension.

Pro Tips

Later transformations can use the variables that were transformed before. This allows you to stack string operations and achieve complex transformations.

Supported transformation

We have several different transformations available now. Please check out transformation reference for more details.

  • replace: Use a regular expression to replace the text in a meta-variable with a new text.
  • substring: Create a new string by cutting off leading and trailing characters.
  • convert: Change the string case of a meta-variable, such as from camelCase to underscore_case.

Add conditional text

Occasionally we may want to add extra text, such as punctuations and newlines, to our fixer string. But whether we should add the new text depends on the presence of absence of other syntax nodes.

A typical scenario is adding a comma between two arguments or list items. We only want to add a comma when the item we are adding is not the last one in the argument list.

We can use replace transformation to create a new meta-variable that only contains text when another meta-variable matches something.

For example, suppose we want to add a new argument to existing function call. We need to add a comma , after the new argument only when the existing call already has some arguments.

id: add-leading-argument
language: python
  pattern: $FUNC($$$ARGS)
      source: $$$ARGS
      replace: '^.+'
      by: ', '

In the above example, if $$$ARGS matches nothing, it will be an empty string and the replace transformation will take no effect. The final fix string will be instantiated to $FUNC(new_argument).

If $$$ARGS does match nodes, then the replacement regular expression will replace the text with ,, so the final fix string will be $FUNC(new_argument, $$$ARGS)

DasSurma Trick

This method is invented by Surma in a tweet, so the useful trick is named after him.

See More in Example Catalog

If you want to see more examples of using ast-grep to rewrite code, you can check out our example catalog. There you can find various use cases and scenarios where ast-grep can help you refactor and improve your code. You can also contribute your own examples and share them with other users.

Made with ❤️ with Rust