Issue #60 - Reduce the memory used to read the response body

[NassimBtk]

Closes #60

What changed

• HttpClient decodes the body straight from the ByteArrayOutputStream buffer with bodyBytes.toString(charset.name()). Before, it copied the buffer with toByteArray() and then decoded the copy. toString(String) exists since Java 1.1 and goes through the same decoder, with the same replacement of malformed input. The charset name comes from Charset.forName(), so it always resolves to the same charset.
• HttpResponse keeps the body as a String instead of a StringBuilder:
  • getBody() returns it without a copy;
  • appendBody() keeps its behavior: it concatenates, and appends "null" for null as StringBuilder.append() did. Each call after the first copies the body; the Javadoc says so, and the client calls it once per response;
  • toString() (header and body) uses concat(), which copies the body once instead of twice.
• Compatibility. The public API does not change, and the library still targets Java 8.

Memory

For an ASCII body of N bytes on JDK 9+, the peak goes from about Cf + 2N to Cf + N, where Cf is the buffer capacity: N with an uncompressed Content-Length, N to 2N when chunked or compressed. In every case one N-byte copy disappears. The benchmark is in #60.

47.7 MiB Kubernetes list	Transfer	1.0.02: min heap	This PR: min heap
ASCII	Content-Length	145 MiB	97 MiB
ASCII	chunked	163 MiB	115 MiB
One char > U+00FF	Content-Length	289 MiB	241 MiB
One char > U+00FF	chunked	307 MiB	259 MiB

The allocation per call is halved (for example 238.7 → 95.5 MiB). With a UTF-16 body, the client now needs no more heap than Files.readString on the same content.

Tests

• New HttpClientTest reads bodies from a local com.sun.net.httpserver server: with a Content-Length, chunked, charset from Content-Type (ISO-8859-1), gzip, empty, and an error status (the body comes from the error stream, and toString() is checked).
• HttpResponseTest also checks appendBody(null), appending "", and that getBody() returns the String without copying it.

mvn verify with JDK 17 (as in CI): 10 tests pass. The Javadoc error about sun.net.www.protocol.http in the build log is not new: main shows it too. The integration tests (httpbin: gzip, deflate, UTF-8, downloadTo, status codes) run in the CI workflow.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-08T11:49:20.960939Z	d075702	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.