arXiv subjects

Juan Altmayer Pizzorno

Publications and source records attributed to Juan Altmayer Pizzorno.

5 recordsLinked to original sources

arXiv2026

Getting Python Types Right with RightTyper

Python type annotations enable static type checking, but most code remains untyped because manual annotation is time-consuming and tedious. Past approaches to automatic type inference fall short: static methods struggle with dynamic features and infer overly broad types; AI-based methods are unsound and miss rare types; and dynamic methods impose extreme overheads (up to 270x), lack important language support such as inferring variable types, or produce annotations that cause runtime errors. This paper presents RightTyper, a novel hybrid approach for Python that produces accurate and precise type annotations grounded in actual program behavior. RightTyper grounds inference in types observed during actual program execution and combines these observations with static analysis and name resolution to produce substantially higher-quality type annotations than prior approaches. Through principled, statistically guided adaptive sampling, RightTyper balances runtime overhead with the need to observe sufficient execution behavior to infer high-quality type annotations. We evaluate RightTyper against static, dynamic, and AI-based systems on both synthetic benchmarks and real-world code, and find that it consistently achieves higher semantic similarity to ground-truth and developer-written annotations, respectively, while incurring only approximately 27% runtime overhead.

cs.PL↗

arXiv2025

CoverUp: Effective High Coverage Test Generation for Python

Testing is an essential part of software development. Test generation tools attempt to automate the otherwise labor-intensive task of test creation, but generating high-coverage tests remains challenging. This paper proposes CoverUp, a novel approach to driving the generation of high-coverage Python regression tests. CoverUp combines coverage analysis, code context, and feedback in prompts that iteratively guide the LLM to generate tests that improve line and branch coverage. We evaluate our prototype CoverUp implementation across a benchmark of challenging code derived from open-source Python projects and show that CoverUp substantially improves on the state of the art. Compared to CodaMosa, a hybrid search/LLM-based test generator, CoverUp achieves a per-module median line+branch coverage of 80% (vs. 47%). Compared to MuTAP, a mutation- and LLM-based test generator, CoverUp achieves an overall line+branch coverage of 89% (vs. 77%). We also demonstrate that CoverUp's performance stems not only from the LLM used but from the combined effectiveness of its components.

cs.SE↗

arXiv2023

SlipCover: Near Zero-Overhead Code Coverage for Python

Coverage analysis is widely used but can suffer from high overhead. This overhead is especially acute in the context of Python, which is already notoriously slow (a recent study observes a roughly 30x slowdown vs. native code). We find that the state-of-the-art coverage tool for Python, coverage$.$py, introduces a median overhead of 180% with the standard Python interpreter. Slowdowns are even more extreme when using PyPy, a JIT-compiled Python implementation, with coverage$.$py imposing a median overhead of 1,300%. This performance degradation reduces the utility of coverage analysis in most use cases, including testing and fuzzing, and precludes its use in deployment. This paper presents SlipCover, a novel, near-zero overhead coverage analyzer for Python. SlipCover works without modifications to either the Python interpreter or PyPy. It first processes a program's AST to accurately identify all branches and lines. SlipCover then dynamically rewrites Python bytecodes to add lightweight instrumentation to each identified branch and line. At run time, SlipCover periodically de-instruments already-covered lines and branches. The result is extremely low overheads -- a median of just 5% -- making SlipCover suitable for use in deployment. We show its efficiency can translate to significant increases in the speed of coverage-based clients. As a proof of concept, we integrate SlipCover into TPBT, a targeted property-based testing system, and observe a 22x speedup.

cs.SE↗

arXiv2023

A Symmetric Dual Encoding Dense Retrieval Framework for Knowledge-Intensive Visual Question Answering

Knowledge-Intensive Visual Question Answering (KI-VQA) refers to answering a question about an image whose answer does not lie in the image. This paper presents a new pipeline for KI-VQA tasks, consisting of a retriever and a reader. First, we introduce DEDR, a symmetric dual encoding dense retrieval framework in which documents and queries are encoded into a shared embedding space using uni-modal (textual) and multi-modal encoders. We introduce an iterative knowledge distillation approach that bridges the gap between the representation spaces in these two encoders. Extensive evaluation on two well-established KI-VQA datasets, i.e., OK-VQA and FVQA, suggests that DEDR outperforms state-of-the-art baselines by 11.6% and 30.9% on OK-VQA and FVQA, respectively. Utilizing the passages retrieved by DEDR, we further introduce MM-FiD, an encoder-decoder multi-modal fusion-in-decoder model, for generating a textual answer for KI-VQA tasks. MM-FiD encodes the question, the image, and each retrieved passage separately and uses all passages jointly in its decoder. Compared to competitive baselines in the literature, this approach leads to 5.5% and 8.5% improvements in terms of question answering accuracy on OK-VQA and FVQA, respectively.

cs.CV↗

arXiv2022

Triangulating Python Performance Issues with Scalene

This paper proposes Scalene, a profiler specialized for Python. Scalene combines a suite of innovations to precisely and simultaneously profile CPU, memory, and GPU usage, all with low overhead. Scalene's CPU and memory profilers help Python programmers direct their optimization efforts by distinguishing between inefficient Python and efficient native execution time and memory usage. Scalene's memory profiler employs a novel sampling algorithm that lets it operate with low overhead yet high precision. It also incorporates a novel algorithm that automatically pinpoints memory leaks, whether within Python or across the Python-native boundary. Scalene tracks a new metric called copy volume, which highlights costly copying operations that can occur when Python silently converts between C and Python data representations, or between CPU and GPU. Since its introduction, Scalene has been widely adopted, with over 500,000 downloads to date. We present experience reports from developers who used Scalene to achieve significant performance improvements and memory savings.

cs.PL↗