pyhtml2text 0.1.0__tar.gz → 0.2.0__tar.gz

pyhtml2text-0.2.0/CHANGELOG.md

@@ -0,0 +1,21 @@
+# Changelog
+
+## 0.2.0 (2026-10-15)
+
+- 1.5–1.7× faster than 0.1.0 on real web pages, and 2× faster on short
+  snippets. Output is unchanged.
+  - Wheels for Linux x86_64/aarch64, macOS and Windows x64 are built with
+    profile-guided optimisation.
+  - Uses mimalloc as the allocator (except on glibc armv7, ppc64le and s390x).
+  - Style rules are shared rather than rebuilt and copied for every call.
+- Deeply nested HTML no longer takes quadratic time to parse: 40,000 nested
+  `<ul>` lists went from 6s to 0.14s (patched html5ever, see `vendor/`).
+- `html` can be UTF-8 `bytes` as well as `str`.
+- An internal error now raises `ValueError` rather than `PanicException`
+  (which `except Exception` doesn't catch).
+- Supports free-threaded Python without re-enabling the GIL.
+- Requires Python 3.10 or later (3.9 is end-of-life).
+
+## 0.1.0 (2026-10-15)
+
+- First release: `html2text()` with `base_url` support.

{pyhtml2text-0.1.0 → pyhtml2text-0.2.0}/Cargo.lock

@@ -14,6 +14,16 @@ version = "2.13.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "3ded4057c258ba199e2d26386d3af3780957ecaee6c4ef4041c6b4b8b97c0b06"
 
+[[package]]
+name = "cc"
+version = "1.8.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6651c9ed80effdc7db0ff72512157f901af5e3549e341e24b1dd4887d836d838"
+dependencies = [
+ "find-msvc-tools",
+ "shlex",
+]
+
 [[package]]
 name = "cfg-if"
 version = "1.0.5"
@@ -26,6 +36,12 @@ version = "2.5.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "da7c62ceae207dd37ea5b845da6a0696c799f85e97da1ab5b7910be3c1c80223"
 
+[[package]]
+name = "find-msvc-tools"
+version = "0.1.14"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "aedcfb3409746eddb02b9e19ebda1c3394f759a152e48ee875a0844d1b955484"
+
 [[package]]
 name = "heck"
 version = "0.5.0"
@@ -35,7 +51,7 @@ checksum = "2304e00983f87ffb38b55b444b5e3b60a884b5d30c0fca7d82fe33449bbe55ea"
 [[package]]
 name = "html2text"
 version = "0.17.1"
-source = "git+https://github.com/jaredlynskey/rust-html2text?rev=ee049006a900891e9df46299004c38ff59c41d65#ee049006a900891e9df46299004c38ff59c41d65"
+source = "git+https://github.com/jaredlynskey/rust-html2text?rev=9c109fac38728040cf3c32b411305cdd840bd511#9c109fac38728040cf3c32b411305cdd840bd511"
 dependencies = [
  "html5ever",
  "tendril",
@@ -46,8 +62,6 @@ dependencies = [
 [[package]]
 name = "html5ever"
 version = "0.39.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "46a1761807faccc9a19e86944bbf40610014066306f96edcdedc2fb714bcb7b8"
 dependencies = [
  "log",
  "markup5ever",
@@ -68,6 +82,15 @@ version = "0.2.190"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ce5d3ddc6d3fa000eb1536d85e147bfe31aacaba692ed6a876f95cb7c855be78"
 
+[[package]]
+name = "libmimalloc-sys"
+version = "0.1.49"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6a45a52f43e1c16f667ccfe4dd8c85b7f7c204fd5e3bf46c5b0db9a5c3c0b8e9"
+dependencies = [
+ "cc",
+]
+
 [[package]]
 name = "lock_api"
 version = "0.4.14"
@@ -103,6 +126,15 @@ dependencies = [
  "autocfg",
 ]
 
+[[package]]
+name = "mimalloc"
+version = "0.1.52"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2d4139bb28d14ad1facf21d5eb8825051b326e172d216b39f6d31df53cc97862"
+dependencies = [
+ "libmimalloc-sys",
+]
+
 [[package]]
 name = "new_debug_unreachable"
 version = "1.0.6"
@@ -200,9 +232,10 @@ dependencies = [
 
 [[package]]
 name = "pyhtml2text"
-version = "0.1.0"
+version = "0.2.0"
 dependencies = [
  "html2text",
+ "mimalloc",
  "pyo3",
 ]
 
@@ -326,6 +359,12 @@ dependencies = [
  "syn 3.0.9",
 ]
 
+[[package]]
+name = "shlex"
+version = "2.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f8fadd59c855ef2080decdef8ff161eb6661b86933c9d82e5ba29dc602a55aba"
+
 [[package]]
 name = "siphasher"
 version = "1.0.4"

pyhtml2text-0.2.0/Cargo.toml

@@ -0,0 +1,32 @@
+[package]
+name = "pyhtml2text"
+version = "0.2.0"
+edition = "2021"
+license = "MIT"
+publish = false
+readme = "README.md"
+
+[lib]
+name = "_pyhtml2text"
+crate-type = ["cdylib"]
+
+[dependencies]
+pyo3 = { version = "0.26", features = ["extension-module", "abi3-py310"] }
+# Fork of jugglerchris/rust-html2text with Config::base_url and some
+# performance changes (branch pyhtml2text-perf).
+html2text = { git = "https://github.com/jaredlynskey/rust-html2text", rev = "9c109fac38728040cf3c32b411305cdd840bd511" }
+
+# Much faster than the system allocators for html2text's many small allocations.
+# Not on glibc armv7/ppc64le/s390x: their manylinux2014 cross compilers are too
+# old to build it (keep in sync with the cfg in src/lib.rs).
+[target.'cfg(not(all(target_env = "gnu", any(target_arch = "arm", target_arch = "powerpc64", target_arch = "s390x"))))'.dependencies]
+mimalloc = { version = "0.1", default-features = false, features = ["local_dynamic_tls"] }
+
+[patch.crates-io]
+# Fixes quadratic parsing of deeply nested HTML, see vendor/README.md.
+html5ever = { path = "vendor/html5ever" }
+
+[profile.release]
+lto = true
+codegen-units = 1
+strip = true

{pyhtml2text-0.1.0 → pyhtml2text-0.2.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pyhtml2text
-Version: 0.1.0
+Version: 0.2.0
 Classifier: Programming Language :: Rust
 Classifier: Programming Language :: Python :: 3
 Classifier: Topic :: Text Processing :: Markup :: HTML
@@ -9,7 +9,7 @@ Summary: Fast HTML to plain text, with relative link resolution (Python bindings
 Keywords: html,text,markdown,html2text
 Author-email: Jared Lynskey <jared@lynskey.co.nz>
 License-Expression: MIT
-Requires-Python: >=3.9
+Requires-Python: >=3.10
 Description-Content-Type: text/markdown; charset=UTF-8; variant=GFM
 Project-URL: Source, https://github.com/jaredlynskey/pyhtml2text
 Project-URL: Upstream, https://github.com/jugglerchris/rust-html2text
@@ -33,6 +33,9 @@ print(html2text(html, base_url="https://example.com/docs/page.html"))
 # [1]: https://example.com/login
 ```
 
+`html` can be a `str` or UTF-8 `bytes` (e.g. `response.content`). Requires
+Python 3.10 or later.
+
 ## Options
 
 `html2text(html, width=80, *, ...)`
@@ -47,10 +50,30 @@ print(html2text(html, base_url="https://example.com/docs/page.html"))
 | `max_wrap_width` | `None` | wrap paragraphs narrower than `width` |
 | `raw` | `False` | flatten tables so cell text stays together |
 
+## Performance
+
+Median time to render a page, on an Apple Silicon Mac:
+
+| | Wikipedia article (780 KB) | BBC News front page (370 KB) |
+|---|---|---|
+| [html2text](https://pypi.org/project/html2text/) (pure Python) | 71.5 ms | 14.9 ms |
+| [html2text-rs](https://pypi.org/project/html2text-rs/) | 15.2 ms | 2.5 ms |
+| **pyhtml2text** | **10.9 ms** | **1.6 ms** |
+
+- The GIL is released while rendering, so threads scale across cores. On
+  free-threaded Python (3.14t) the module runs without re-enabling the GIL.
+- Wheels for Linux (x86_64, aarch64), macOS and Windows x64 are built with
+  profile-guided optimisation, which makes them about 25% faster. Other
+  platforms get the same code without it.
+- Deeply nested HTML is parsed in linear time, so hostile input can't stall a
+  worker: 40,000 nested lists take 0.14 s, against 6 s with stock html5ever.
+
 ## Credits
 
 The rendering is done by rust-html2text by Chris Emerson (MIT). This package
-uses [a fork](https://github.com/jaredlynskey/rust-html2text/tree/base-url-resolution)
-that adds `base_url`. It isn't affiliated with or endorsed by the upstream project,
-so please report issues here rather than upstream.
+uses [a fork](https://github.com/jaredlynskey/rust-html2text/tree/pyhtml2text-perf)
+that adds `base_url` and some performance changes, and a patched copy of
+[html5ever](https://github.com/servo/html5ever) (MIT/Apache-2.0, see
+`vendor/`). It isn't affiliated with or endorsed by either upstream project, so
+please report issues here rather than upstream.
 

{pyhtml2text-0.1.0 → pyhtml2text-0.2.0}/README.md

@@ -17,6 +17,9 @@ print(html2text(html, base_url="https://example.com/docs/page.html"))
 # [1]: https://example.com/login
 ```
 
+`html` can be a `str` or UTF-8 `bytes` (e.g. `response.content`). Requires
+Python 3.10 or later.
+
 ## Options
 
 `html2text(html, width=80, *, ...)`
@@ -31,9 +34,29 @@ print(html2text(html, base_url="https://example.com/docs/page.html"))
 | `max_wrap_width` | `None` | wrap paragraphs narrower than `width` |
 | `raw` | `False` | flatten tables so cell text stays together |
 
+## Performance
+
+Median time to render a page, on an Apple Silicon Mac:
+
+| | Wikipedia article (780 KB) | BBC News front page (370 KB) |
+|---|---|---|
+| [html2text](https://pypi.org/project/html2text/) (pure Python) | 71.5 ms | 14.9 ms |
+| [html2text-rs](https://pypi.org/project/html2text-rs/) | 15.2 ms | 2.5 ms |
+| **pyhtml2text** | **10.9 ms** | **1.6 ms** |
+
+- The GIL is released while rendering, so threads scale across cores. On
+  free-threaded Python (3.14t) the module runs without re-enabling the GIL.
+- Wheels for Linux (x86_64, aarch64), macOS and Windows x64 are built with
+  profile-guided optimisation, which makes them about 25% faster. Other
+  platforms get the same code without it.
+- Deeply nested HTML is parsed in linear time, so hostile input can't stall a
+  worker: 40,000 nested lists take 0.14 s, against 6 s with stock html5ever.
+
 ## Credits
 
 The rendering is done by rust-html2text by Chris Emerson (MIT). This package
-uses [a fork](https://github.com/jaredlynskey/rust-html2text/tree/base-url-resolution)
-that adds `base_url`. It isn't affiliated with or endorsed by the upstream project,
-so please report issues here rather than upstream.
+uses [a fork](https://github.com/jaredlynskey/rust-html2text/tree/pyhtml2text-perf)
+that adds `base_url` and some performance changes, and a patched copy of
+[html5ever](https://github.com/servo/html5ever) (MIT/Apache-2.0, see
+`vendor/`). It isn't affiliated with or endorsed by either upstream project, so
+please report issues here rather than upstream.

{pyhtml2text-0.1.0 → pyhtml2text-0.2.0}/pyproject.toml

@@ -4,12 +4,12 @@ build-backend = "maturin"
 
 [project]
 name = "pyhtml2text"
-version = "0.1.0"
+version = "0.2.0"
 description = "Fast HTML to plain text, with relative link resolution (Python bindings to rust-html2text)"
 readme = "README.md"
 license = "MIT"
 license-files = ["LICENSE"]
-requires-python = ">=3.9"
+requires-python = ">=3.10"
 authors = [{ name = "Jared Lynskey", email = "jared@lynskey.co.nz" }]
 keywords = ["html", "text", "markdown", "html2text"]
 classifiers = [
@@ -25,6 +25,19 @@ Upstream = "https://github.com/jugglerchris/rust-html2text"
 [tool.maturin]
 python-source = "python"
 module-name = "pyhtml2text._pyhtml2text"
+# cargo leaves out vendor/html5ever (it has its own Cargo.toml), but the
+# [patch] in Cargo.toml needs it; the PGO data and CI files aren't needed.
+include = [
+  { path = "vendor/html5ever/Cargo.toml", format = "sdist" },
+  { path = "vendor/html5ever/LICENSE-*", format = "sdist" },
+  { path = "vendor/html5ever/README.md", format = "sdist" },
+  { path = "vendor/html5ever/src/**/*.rs", format = "sdist" },
+]
+exclude = [
+  { path = "pgo/**/*", format = "sdist" },
+  { path = "bench/**/*", format = "sdist" },
+  { path = ".github/**/*", format = "sdist" },
+]
 
 [tool.maturin.generate-ci.github]
 pytest = true

{pyhtml2text-0.1.0 → pyhtml2text-0.2.0}/python/pyhtml2text/__init__.pyi

@@ -1,14 +1,12 @@
-from typing import Optional
-
 def html2text(
-    html: str,
+    html: str | bytes,
     width: int = 80,
     *,
-    base_url: Optional[str] = None,
+    base_url: str | None = None,
     decorate: bool = True,
     link_footnotes: bool = True,
     table_borders: bool = True,
     wrap_links: bool = True,
-    max_wrap_width: Optional[int] = None,
+    max_wrap_width: int | None = None,
     raw: bool = False,
 ) -> str: ...

pyhtml2text-0.2.0/src/lib.rs

@@ -0,0 +1,99 @@
+use std::panic::{catch_unwind, AssertUnwindSafe};
+
+use html2text::config;
+use pyo3::exceptions::{PyTypeError, PyValueError};
+use pyo3::prelude::*;
+use pyo3::types::{PyBytes, PyString};
+
+#[cfg(not(all(
+    target_env = "gnu",
+    any(target_arch = "arm", target_arch = "powerpc64", target_arch = "s390x")
+)))]
+#[global_allocator]
+static GLOBAL: mimalloc::MiMalloc = mimalloc::MiMalloc;
+
+/// Render HTML as plain text.
+#[pyfunction]
+#[pyo3(name = "html2text", signature = (
+    html,
+    width = 80,
+    *,
+    base_url = None,
+    decorate = true,
+    link_footnotes = true,
+    table_borders = true,
+    wrap_links = true,
+    max_wrap_width = None,
+    raw = false,
+))]
+#[allow(clippy::too_many_arguments)]
+fn render(
+    py: Python<'_>,
+    html: &Bound<'_, PyAny>,
+    width: usize,
+    base_url: Option<String>,
+    decorate: bool,
+    link_footnotes: bool,
+    table_borders: bool,
+    wrap_links: bool,
+    max_wrap_width: Option<usize>,
+    raw: bool,
+) -> PyResult<String> {
+    // Borrow the input without copying it; bytes are taken to be UTF-8.
+    let input: &[u8] = if let Ok(s) = html.downcast::<PyString>() {
+        s.to_str()?.as_bytes()
+    } else if let Ok(b) = html.downcast::<PyBytes>() {
+        b.as_bytes()
+    } else {
+        return Err(PyTypeError::new_err(format!(
+            "html must be str or bytes, not {}",
+            html.get_type().name()?
+        )));
+    };
+
+    let mut conf = if decorate {
+        config::plain()
+    } else {
+        config::plain_no_decorate()
+    };
+    conf = conf.link_footnotes(link_footnotes).allow_width_overflow();
+    if let Some(base) = base_url {
+        conf = conf.base_url(base);
+    }
+    if !table_borders {
+        conf = conf.no_table_borders();
+    }
+    if !wrap_links {
+        conf = conf.no_link_wrapping();
+    }
+    if let Some(w) = max_wrap_width {
+        conf = conf.max_wrap_width(w);
+    }
+    if raw {
+        conf = conf.raw_mode(true);
+    }
+
+    // Rendering is pure Rust, so let other Python threads run meanwhile. A
+    // panic would otherwise surface as PanicException, which `except
+    // Exception` doesn't catch, so report it as a ValueError instead.
+    match py.detach(|| catch_unwind(AssertUnwindSafe(|| conf.string_from_read(input, width)))) {
+        Ok(Ok(text)) => Ok(text),
+        Ok(Err(e)) => Err(PyValueError::new_err(e.to_string())),
+        Err(panic) => {
+            let msg = panic
+                .downcast_ref::<&str>()
+                .copied()
+                .or_else(|| panic.downcast_ref::<String>().map(String::as_str))
+                .unwrap_or("unknown error");
+            Err(PyValueError::new_err(format!(
+                "html2text failed on this document: {msg}"
+            )))
+        }
+    }
+}
+
+#[pymodule(gil_used = false)]
+fn _pyhtml2text(m: &Bound<'_, PyModule>) -> PyResult<()> {
+    m.add_function(wrap_pyfunction!(render, m)?)?;
+    Ok(())
+}

{pyhtml2text-0.1.0 → pyhtml2text-0.2.0}/tests/test_basic.py

@@ -56,3 +56,34 @@ def test_tables():
 def test_unicode_and_empty():
     assert html2text("<p>한국어 텍스트 é</p>") == "한국어 텍스트 é\n"
     assert html2text("") == ""
+
+
+def test_bytes_input():
+    assert html2text("<p>café</p>".encode()) == "café\n"
+    assert html2text(b'<a href="/x">x</a>', base_url=PAGE) == html2text(
+        '<a href="/x">x</a>', base_url=PAGE
+    )
+
+
+def test_rejects_other_types():
+    import pytest
+
+    with pytest.raises(TypeError, match="str or bytes"):
+        html2text(123)
+
+
+def test_deep_nesting_is_linear():
+    # Parsing used to be quadratic in nesting depth (40,000 nested lists took
+    # 6s); with the patched html5ever this takes well under a second.
+    import platform
+    import sys
+    import time
+
+    html = "<ul><li>" * 100_000 + "x"
+    start = time.perf_counter()
+    out = html2text(html)
+    elapsed = time.perf_counter() - start
+    assert out.rstrip().endswith("x")
+    # Only time it where tests run natively; emulated CI runners are slow.
+    if platform.machine().lower() in ("x86_64", "amd64") or sys.platform == "darwin":
+        assert elapsed < 10, f"took {elapsed:.1f}s"

pyhtml2text-0.2.0/vendor/README.md

@@ -0,0 +1,13 @@
+# Vendored crates
+
+## html5ever 0.39.0
+
+Copied from the crates.io release (`src/`, licences, README and manifest;
+examples, tests and benchmarks removed). Dual-licensed MIT / Apache-2.0,
+see `html5ever/LICENSE-*`.
+
+Patched (see git history of this directory) so that checking for an open `<p>`
+element "in button scope" is O(1) instead of walking the whole stack of open
+elements. That check runs for most block-level start tags, so deeply nested
+input (e.g. tens of thousands of nested `<ul>`/`<div>`/`<blockquote>`) took
+quadratic time to parse. Parse results are unchanged.

pyhtml2text-0.2.0/vendor/html5ever/Cargo.toml

@@ -0,0 +1,61 @@
+# THIS FILE IS AUTOMATICALLY GENERATED BY CARGO
+#
+# When uploading crates to the registry Cargo will automatically
+# "normalize" Cargo.toml files for maximal compatibility
+# with all versions of Cargo and also rewrite `path` dependencies
+# to registry (e.g., crates.io) dependencies.
+#
+# If you are reading this file be aware that the original Cargo.toml
+# will likely look very different (and much more reasonable).
+# See Cargo.toml.orig for the original contents.
+
+[package]
+edition = "2021"
+rust-version = "1.71.0"
+name = "html5ever"
+version = "0.39.0"
+authors = ["The html5ever Project Developers"]
+build = false
+autolib = false
+autobins = false
+autoexamples = false
+autotests = false
+autobenches = false
+description = "High-performance browser-grade HTML5 parser"
+documentation = "https://docs.rs/html5ever"
+readme = "README.md"
+keywords = [
+    "html",
+    "html5",
+    "parser",
+    "parsing",
+]
+categories = [
+    "parser-implementations",
+    "web-programming",
+]
+license = "MIT OR Apache-2.0"
+repository = "https://github.com/servo/html5ever"
+
+[features]
+serde = ["markup5ever/serde"]
+trace_tokenizer = []
+
+[lib]
+name = "html5ever"
+path = "src/lib.rs"
+
+
+
+
+
+
+
+
+[dependencies.log]
+version = "0.4"
+
+[dependencies.markup5ever]
+version = "0.39"
+
+