AWS DevOps & Developer Productivity Blog

Five ways to optimize code with Amazon Q Developer

Practical improvement and optimization of software quality requires expert-level knowledge across various subjects. As such, in this blog we shall look at how Amazon Q Developer can help improve your development team productivity and application stability by enabling automation around code optimization by improving your code’s quality, performance, application infrastructure specifications.

The blog will also look at sample prompts that can be used to discover optimization options, control the scope of modifications, choose improvements and iterate through code changes. Being a generative AI–powered software development assistant that integrates with your integrated development environment (IDE), Amazon Q Developer supports in code explanation, code generation, and code improvements such as debugging and optimization. Amazon Q Developer can be configured for IDEs such as Visual Studio Code or Jet Brains IDEs, using AWS Identity and Access Management (IAM) Identity Center or AWS Builder ID.

To illustrate the optimization techniques, we will use the quant-trading sample application from the github aws-samples repo, to look at optimizations across the following domains – 1) Portability 2) Complexity 3) Code Performance 4) Infrastructure 5) Architecture and non-functionals 6) Running on AWS

Please note that as Amazon Q Developer continues to evolve, and due to the non-deterministic nature of Generative AI, the outputs you see when trying this yourself may differ from the examples shown in this blog post.

Amazon Q Developer can assess your code, provide recommendations, and generate an optimized version based on your prompts. A prompt is a natural language text that requests the generative AI to perform a specific task. Among areas you can optimize are portability and complexity.

Portability optimization

To assess portability of your code base, Let us use Portfolio Generator python code from quant-trading sample.

  • In the Integrated development environment (IDE), select the entire code in the file, open Amazon Q Chat and type your prompt: “Is the selected code portable?”

Amazon Q Developer will generate an assessment of portability of your code, as shown in Figure 1. Any specific improvements possible will also be specified.

This image shows two side-by-side screenshots of an Amazon Q chat interface discussing code portability. The left panel displays a question "Is the selected code portable?" followed by a detailed response outlining factors affecting code portability, including use of relative imports, hard-coded paths, external libraries, and AWS SDK integration. The right panel continues the discussion with suggestions on how to make the code more portable, including using absolute imports, avoiding hard-coded paths, isolating dependencies, and separating AWS-specific functionality. The interface has a dark theme with white text on a black background. At the bottom, there are suggested follow-up questions and a note about the AWS Responsible AI Policy.

Figure 1: Optimize code quality. Assessment and recommendations

  • Add code snippets directly to the prompt as context, for further response improvements by:
    1. Right click on the IDE
    2. choose “Send to Amazon Q”
    3. Select “Send to Prompt”.

Now, the context includes the code, its portability assessment and recommendations for further improvements.

  • Ask – “Rewrite code for maximum portability”

However, such a generic prompt would likely result in numerous code modifications chosen by Amazon Q Developer, as shown in Figure 2. To achieve a more specific and higher quality output, in addition to enriched context, the prompt must be more precise and targeted.

This image shows three side-by-side panels of code and explanations in an Amazon Q chat interface. The left panel displays Python code with various import statements and function definitions. The middle panel contains a summary of key changes made to improve code portability, including the use of environment variables, absolute imports, argument parsing, decimal usage, error handling, and formatting. The right panel shows more detailed Python code with import statements, function definitions, and file path configurations. All panels have a dark theme with light-colored text on a dark background. The interface includes options for asking questions or accessing quick actions at the bottom of each panel.

Figure 2: Specific optimization – externalizing config.

  • Ask Amazon Q Developer to perform optimization addressing only hardcoded path values in a specific way.
    • “Rewrite this code to be more portable. Move hardcoded file paths into a separate JSON configuration file under the node “file-paths”. Leave the rest of the file unchanged.”

Amazon Q Developer will now rewrite a few lines of the code and externalized configuration into a JSON file, as shown in Figure 3.

This image shows three panels of an Amazon Q chat interface discussing code portability improvement. The left panel displays a request to rewrite code for better portability by moving hardcoded file paths to a JSON configuration file. It then shows the rewritten Python code with import statements and a highlighted section for loading file paths from the configuration file. The middle panel contains some Python code and an example of the JSON configuration file with "file-paths" node. It explains how the rewritten code loads file paths from the config.json file, making the code more portable and easier to modify for different environments. The right panel shows more detailed Python code, including import statements and function definitions. A section of this code is highlighted, showing sys.path.append() statements that are likely the target of the portability improvement. All panels have a dark theme with colorful syntax highlighting for the code. The interface includes options for asking questions or accessing quick actions at the bottom of each panel.

    Figure 3: Specific optimization – externalizing config.

Note: Dialogue with Amazon Q Developer can span several iterations, allowing you to analyze and narrow down to a very specific aspect of your code. This approach will appear in line with pair programming, iteratively collaborating on a better solution.

  • Continue iterating for optimizations per your code. Examples are – ask “Use YAML format for config.” or “Use path names in config similar to their original values.” or “Add error handling when working with files.”

Such an iterative approach will allow you to gradually apply modifications while preserving control over the scope of changes.

Complexity Optimization

Now let’s analyze and reduce the complexity of the write_portfolio method:

  1. Ask either:
    • “Can the selected code be simplified?”
    • “How can I reduce complexity of the selected code?”
  2. Drill down into a specific, scoped optimization.
    • “Simplify loops, conditions and variables of the selected code.”

Be specific about the kind of optimizations you want Amazon Q Developer to apply (see Figure 4). Example, ask direct prompts such as – “Replace portfolio dictionary with JSON.”

This image shows two panels of an Amazon Q chat interface discussing code simplification. The left panel displays a request to "Simplify loops, conditions and variables of the selected code" followed by a simplified Python function called write_portfolio. The function creates a portfolio dictionary with various keys and values, and includes simplified logic for selecting tickers and creating a positions list using list comprehension. The right panel shows the original Python code that is being simplified. This code includes the write_portfolio function definition with similar structure but more verbose implementation. The file path at the top indicates this is from a file named portfolio_generator.py. Both panels use a dark theme with syntax highlighting in various colors for better code readability. The interface includes an option to ask questions or enter commands at the bottom of the left panel.

Figure 4: Simplify code example

Code Performance optimization

To improve code performance, we shall leverage Amazon Q Developer’s “Optimize” feature. It initiates a dialogue for code performance optimization via the right-click menu or key shortcut (see Figure 5).

This image shows two main panels of an Amazon Q chat interface discussing code optimization. The left panel displays a request to optimize a specific part of the code, followed by suggestions for improvement. These suggestions include using generator expressions instead of list comprehensions, avoiding unnecessary conversions, using conditional assignment, and considering NumPy or Pandas for large numerical datasets. Each suggestion is accompanied by a code snippet demonstrating the optimization. The right panel shows the original Python code in a file editor, with the function calculate_weights highlighted. This appears to be the function targeted for optimization. The editor interface includes various options like "Go to Definition", "Find All References", and "Optimize" visible in a dropdown menu. Both panels use a dark theme with syntax highlighting in various colors for better code readability. The interface includes tabs at the top for different files or chat sessions, and an option to ask questions or enter commands at the bottom of the left panel.

Figure 5: IDE “built-in” feature for code improvement. Amazon Q -> Optimize

The selected code is sent to Amazon Q Developer, which then provides recommendations and generates optimized code.

Let’s now look at how we can use Amazon Q Developer to improve the calculate_weights method.

As shown in Figure 5, Amazon Q Developer explains step-by-step every optimization it suggests. You can further follow-up with a more precise prompt, targeting a specific optimization for a specific code block. For instance, “Optimize only selected method and only avoid unnecessary type conversions. Leave the rest of code unchanged.”

A screenshot of a code editor displaying Python code with a dark background theme. The image shows multiple functions and methods, including 'calculate_weights', 'get_final_payload', and 'add_parameter'. On the left side, there's a blue banner with instructions to optimize a selected method and avoid unnecessary type conversions. Below this, an explanation of the optimized 'calculate_weights' method is provided, highlighting changes made to improve performance. The code is syntax-highlighted, making different elements like functions, variables, and comments easily distinguishable.

Figure 6: Follow-up with a more specific prompt for performance optimization

You can copy-paste newly generated code or insert it directly at the cursor by choosing “Insert code”.

To achieve even higher precision, include in your prompt what not to do or to avoid.

Infrastructure optimization

Amazon Q Developer also supports Infrastructure as Code (IaC) out of the box, providing expert advice and code generation for CloudFormation, CDK, and Terraform. This allows you to leverage code optimization techniques and patterns for your infrastructure.

As a demonstration, let’s improve portability of the CDK code in lambda.ts by introducing environment variables to inject configurations into the runtime.

To begin,

  1. Start a new chat with a broad question – “Could you recommend techniques to inject system variables into a Lambda container function?” Amazon Q Developer will generally provide options to inject environment variables into an AWS Lambda function.
  2. Send function code to the prompt and ask Amazon Q Developer. This generates the code for injecting environment variables through Lambda runtime by using prompt – “Could you add some deployment variables into the tradingStartStopFunction function?”
This image shows three side-by-side screenshots of an Amazon Q chat interface and code editor. The left panel displays a conversation about injecting system variables into a Lambda container function, listing five techniques. The middle panel shows a code snippet for a 'tradingStartStopFunction' with a question about adding deployment variables. The right panel displays more detailed code for Lambda functions related to trading operations. All three panels have a dark theme with syntax-highlighted code in various colors.

Figure 7: Optimizing infrastructure code by introducing environment variables in a Lambda function

Architecture and non-functional optimization

With Amazon Q Developer, you can go beyond code and enhance your system architecture. Let’s consider lambda_function.py which interacts with Amazon DynamoDB and AWS Systems Manager Parameter Store.

  • Send the entire function to the prompt and ask the following in sequence.
    • “What are the architecture implications if I call this lambda function daily?”
    • “How do I optimize this function to be called daily.”
    • Then, follow up with –“How do I optimize this function to be called every 1 second.”
A split-screen image showing two chat conversations and a code editor. The left panel discusses architectural implications of calling a Lambda function daily, covering topics like concurrency, idempotency, error handling, separation of concerns, monitoring, and security. The middle panel offers optimization strategies for calling a Lambda function every 1 second, including separating concerns, caching, batching, and scaling. The right panel shows Python code for a Lambda function, including imports and a function definition dealing with DynamoDB operations.A split-screen image showing two chat conversations and a code editor. The left panel discusses architectural implications of calling a Lambda function daily, covering topics like concurrency, idempotency, error handling, separation of concerns, monitoring, and security. The middle panel offers optimization strategies for calling a Lambda function every 1 second, including separating concerns, caching, batching, and scaling. The right panel shows Python code for a Lambda function, including imports and a function definition dealing with DynamoDB operations.

Figure 8: NFRs and business rules impact architecture enhancements

  • Compare Amazon Q’s outputs to see how each use case impacts the architectural recommendations, such as introducing caching, batch processing, queues, or concurrency mechanisms.

Following the techniques discussed earlier, you can dive in more specific implementations of suggested architecture enhancements. For example, ask “Implement a mechanism to execute only one instance of lambda function at any given moment of time. Implement cache for SSM Parameter store value, but not for Portfolio table.”

Optimize code to run on AWS

As a versatile developer assistant, Amazon Q Developer excels at helping you adhere to AWS best practices and recommendations.

Let’s examine if our sample – IntradayMomentum Lambda function handler can be improved.

  • Send the code to the Amazon Q Developer prompt and ask – “Is this lambda handler following AWS recommended best practices?”
This image shows a split-screen view of an Amazon Q chat interface on the left and a code editor on the right. The left side displays a conversation about AWS Lambda function best practices, listing 9 points of improvement for the provided code, including separation of concerns, environment variables usage, logging, error handling, dependency management, performance optimization, security, idempotency, and testing. The right side shows Python code for a Lambda function. The code includes a lambda_handler function with various operations like getting symbols, calculating updates and weights, and interacting with a DynamoDB table. The code is syntax-highlighted, indicating it's being viewed in a code editor. At the top of the code editor, there are tab names suggesting multiple files are open, including "lambda_function.py" and "portfolio_generator.py". The overall theme of the interface is dark, suggesting a dark mode IDE or development environment.

Figure 9: Optimize code to run on AWS. AWS-recommended best practices for the Lambda handler

The analysis generated by Amazon Q Developer is based on AWS code, best practices and documentation. Not only does it suggest improvements, but also highlights what’s been done correctly, reinforcing best practices.

  • Following an iterative technique described earlier, continue asking Amazon Q developer for further recommendations with more specific prompts. For example – “Add exception handling to the code.”
This image shows a split-screen interface with an Amazon Q chat on the left and a code editor on the right, both using a dark theme. The left side displays a chat conversation about adding exception handling to the code. It shows Python code for a Lambda function with newly added exception handling, including imports for logging and a try-except block. The right side shows the original Python code for the Lambda function in a code editor. The code includes functions for handling portfolio updates, interacting with DynamoDB, and processing various data elements. At the top of the screen, there are multiple tabs open in the code editor, including "lambda_function.py", "portfolio_generator.py", and "deploy_portfolio.py". The image demonstrates the process of improving the Lambda function code by adding error handling based on the chat conversation's recommendations.

Figure 10: Rewrite code with Best Practices in place. Adding Exception Handling.

Conclusion

In this blog post, we discussed approaches for code optimization with the help of Amazon Q Developer. We explored code optimization from various perspectives, such as code quality, performance, application infrastructure, following best practices, and enhancing architecture. We saw the importance of prompt engineering and context when optimizing code with Amazon Q Developer – a generative AI coding assistant. Starting with open, generic prompts helps build the necessary context and discover optimization options. In contrast, precise and specific follow-up prompts help define the scope of changes and incrementally generate optimized code.

It has never been easier for developers to have a development assistant and start improving code with the help of natural language dialogue, provided by Amazon Q.

About the authors

Roman Martynenko is a Senior Solutions Architect at Amazon Web Services with over 20 years of experience in Software Engineering, Architecture and Cloud technologies. Roman is helping Canadian public sector customers with their cloud journey. He focuses on next-generation developer experience, helping organizations re-imagine the entire Software Development Lifecycle. Outside of work, he enjoys hiking, home automation, and DIY projects.

Karthik Chemudupati is a Principal Technical Account Manager (TAM) with AWS, focused on helping customers achieve cost optimization and operational excellence. He has more than 20 years of IT experience in software engineering, cloud operations and automations. Karthik joined AWS in 2016 as a TAM and worked with more than dozen Enterprise Customers across US-West. Outside of work, he enjoys spending time with his family.

Shardul Vaidya is a Worldwide Partner Solutions Architect with AWS, focused on helping partners and customers build and effectively use Generative AI powered developer experiences. Shardul joined AWS in 2020 as part of their early career talent Solutions Architect team and worked with over a hundred modernization and DevOps partners across the world. Outside of work, he’s a music lover and collects records.