Repository navigation
Releases
v0.5.0
Blitz-based renderer CLI, built from d8f4b689dfe74e3673f67d2750b000311bc7866c.
Dynamically linked, not static: it names six libraries directly
(libc, libssl, libcrypto, libfontconfig, libm,
libgcc_s) and pulls freetype, expat, libpng, bzip2, brotli and
zlib through fontconfig, so ldd lists 12 of them on bookworm and
13 on jammy.
| Asset | pageplate_0.5.0_linux_amd64.tar.gz |
| Binary | pageplate, 26M, single file |
| Runtime deps | six direct, six more via fontconfig; ldd lists 12 on bookworm, 13 on jammy |
| glibc floor | 2.35 (asserted at build time) |
| Built on | ubuntu-22.04 |
| Blitz pin | see Cargo.toml at this tag |
Install:
V=v0.5.0
curl -fsSL -O "https://github.com/leshchenko1979/pageplate/releases/download/$V/pageplate_0.5.0_linux_amd64.tar.gz"
sha256sum -c <(curl -fsSL "https://github.com/leshchenko1979/pageplate/releases/download/$V/checksums.txt")
tar xzf "pageplate_0.5.0_linux_amd64.tar.gz"
install -m755 pageplate /usr/local/bin/pageplateNot a browser: renders JS-free HTML/CSS to PNG.
Measured by a probe in this repo, not quoted from a private deck:
tools/bench.py renders a seeded eight-slide deck through both
engines on a GitHub runner and writes bench/results.json, which is
committed unedited. Against headless Chromium, peak process-tree RSS
is at least 31x smaller: 75.4 MiB against 2380.2 MiB for
--clip, 47.5 MiB for --clip --per-element, 37.8 MiB for
--full. Wall clock is measured and drift-gated but not advertised,
because the browser's own wall moves 18-22% between runs and a ratio
would mostly report how fast the browser happened to be that day.
4.7102% of pixels differ by more than 16 levels (glyph rasterisation and
antialiasing; how much is layout is not quantified). On one host, same
font stack, renderer version and band budget, three renders of the same input are byte-identical.
Reproduce it:
python3 tools/bench.py --bin <pageplate> --chrome <chrome>.
v0.4.0
Blitz-based renderer CLI, built from 1075001bc7c7cbe5c4053fd5bba39bc924d5f042.
Dynamically linked, not static: it names six libraries directly
(libc, libssl, libcrypto, libfontconfig, libm,
libgcc_s) and pulls freetype, expat, libpng, bzip2, brotli and
zlib through fontconfig, so ldd lists 12 of them on bookworm and
13 on jammy.
| Asset | pageplate_0.4.0_linux_amd64.tar.gz |
| Binary | pageplate, 25M, single file |
| Runtime deps | six direct, six more via fontconfig; ldd lists 12 on bookworm, 13 on jammy |
| glibc floor | 2.35 (asserted at build time) |
| Built on | ubuntu-22.04 |
| Blitz pin | see Cargo.toml at this tag |
Install:
V=v0.4.0
curl -fsSL -O "https://github.com/leshchenko1979/pageplate/releases/download/$V/pageplate_0.4.0_linux_amd64.tar.gz"
sha256sum -c <(curl -fsSL "https://github.com/leshchenko1979/pageplate/releases/download/$V/checksums.txt")
tar xzf "pageplate_0.4.0_linux_amd64.tar.gz"
install -m755 pageplate /usr/local/bin/pageplateNot a browser: renders JS-free HTML/CSS to PNG.
Measured by a probe in this repo, not quoted from a private deck:
tools/bench.py renders a seeded eight-slide deck through both
engines on a GitHub runner and writes bench/results.json, which is
committed unedited. Against headless Chromium, peak process-tree RSS
is at least 31x smaller: 75.4 MiB against 2380.2 MiB for
--clip, 47.5 MiB for --clip --per-element, 37.8 MiB for
--full. Wall clock is measured and drift-gated but not advertised,
because the browser's own wall moves 18-22% between runs and a ratio
would mostly report how fast the browser happened to be that day.
4.7102% of pixels differ by more than 16 levels (glyph rasterisation and
antialiasing; how much is layout is not quantified). On one host, same
font stack, renderer version and band budget, three renders of the same input are byte-identical.
Reproduce it:
python3 tools/bench.py --bin <pageplate> --chrome <chrome>.
v0.3.0-beta.9
Blitz-based renderer CLI, built from 35aba3e4f61a118b4717eb790d8081c18e6e3d4f.
Dynamically linked, not static: the table below lists every shared
library it needs, and ldd on the asset prints the same list.
| Asset | html-render_0.3.0-beta.9_linux_amd64.tar.gz |
| Binary | html-render, 25M, single file |
| Runtime deps | 13 system libraries: glibc, libssl/libcrypto, freetype, fontconfig, expat, png, zlib, bz2, brotli |
| glibc floor | 2.35 (asserted at build time) |
| Built on | ubuntu-22.04 |
| Blitz pin | see Cargo.toml at this tag |
Install:
V=v0.3.0-beta.9
curl -fsSL -O "https://github.com/leshchenko1979/html-render/releases/download/$V/html-render_0.3.0-beta.9_linux_amd64.tar.gz"
sha256sum -c <(curl -fsSL "https://github.com/leshchenko1979/html-render/releases/download/$V/checksums.txt")
tar xzf "html-render_0.3.0-beta.9_linux_amd64.tar.gz"
install -m755 html-render /usr/local/bin/html-renderNot a browser: renders JS-free HTML/CSS to PNG.
Measured on a 12-slide, 4 MB deck at 2560x1440: 5.37 s against
headless Chromium's 13.36 s (2.49x faster). The RSS figures published
for that deck — 93.4 MiB against 858 MiB — are largest single
process, not the process tree; measured on a deck in the repo,
Chromium's tree peaks 5.7-6.1x above its largest process while
html-render is one process, so the real memory gap is wider than the
9.19x those two numbers give. 4.13% of pixels differ by more than 16
levels (glyph antialiasing, not layout), and ten renders of the same
input are byte-identical. Source of these figures: GATES.md.
v0.3.0-beta.8
Blitz-based renderer CLI, built from d24ff5c76922aba16e01e6b655865f9d9193f34d.
Dynamically linked, not static: the table below lists every shared
library it needs, and ldd on the asset prints the same list.
| Asset | html-render_0.3.0-beta.8_linux_amd64.tar.gz |
| Binary | html-render, 25M, single file |
| Runtime deps | 13 system libraries: glibc, libssl/libcrypto, freetype, fontconfig, expat, png, zlib, bz2, brotli |
| glibc floor | 2.35 (asserted at build time) |
| Built on | ubuntu-22.04 |
| Blitz pin | see Cargo.toml at this tag |
Install:
V=v0.3.0-beta.8
curl -fsSL -O "https://github.com/leshchenko1979/html-render/releases/download/$V/html-render_0.3.0-beta.8_linux_amd64.tar.gz"
sha256sum -c <(curl -fsSL "https://github.com/leshchenko1979/html-render/releases/download/$V/checksums.txt")
tar xzf "html-render_0.3.0-beta.8_linux_amd64.tar.gz"
install -m755 html-render /usr/local/bin/html-renderNot a browser: renders JS-free HTML/CSS to PNG.
Measured on a 12-slide, 4 MB deck at 2560x1440: 5.37 s against
headless Chromium's 13.36 s (2.49x faster). The RSS figures published
for that deck — 93.4 MiB against 858 MiB — are largest single
process, not the process tree; measured on a deck in the repo,
Chromium's tree peaks 5.7-6.1x above its largest process while
html-render is one process, so the real memory gap is wider than the
9.19x those two numbers give. 4.13% of pixels differ by more than 16
levels (glyph antialiasing, not layout), and ten renders of the same
input are byte-identical. Source of these figures: GATES.md.
v0.3.0-beta.10
Blitz-based renderer CLI, built from 5e59bdeb10728a608acfca856610c2009c4c6907.
Dynamically linked, not static: the table below lists every shared
library it needs, and ldd on the asset prints the same list.
| Asset | html-render_0.3.0-beta.10_linux_amd64.tar.gz |
| Binary | html-render, 25M, single file |
| Runtime deps | 13 system libraries: glibc, libssl/libcrypto, freetype, fontconfig, expat, png, zlib, bz2, brotli |
| glibc floor | 2.35 (asserted at build time) |
| Built on | ubuntu-22.04 |
| Blitz pin | see Cargo.toml at this tag |
Install:
V=v0.3.0-beta.10
curl -fsSL -O "https://github.com/leshchenko1979/html-render/releases/download/$V/html-render_0.3.0-beta.10_linux_amd64.tar.gz"
sha256sum -c <(curl -fsSL "https://github.com/leshchenko1979/html-render/releases/download/$V/checksums.txt")
tar xzf "html-render_0.3.0-beta.10_linux_amd64.tar.gz"
install -m755 html-render /usr/local/bin/html-renderNot a browser: renders JS-free HTML/CSS to PNG.
Measured on a 12-slide, 4 MB deck at 2560x1440: 5.37 s against
headless Chromium's 13.36 s (2.49x faster). The RSS figures published
for that deck — 93.4 MiB against 858 MiB — are largest single
process, not the process tree; measured on a deck in the repo,
Chromium's tree peaks 5.7-6.1x above its largest process while
html-render is one process, so the real memory gap is wider than the
9.19x those two numbers give. 4.13% of pixels differ by more than 16
levels (glyph antialiasing, not layout), and ten renders of the same
input are byte-identical. Source of these figures: GATES.md.
v0.3.0-beta.7
Blitz-based renderer CLI, built from aaa26bae5ce214f2216348a1f8d27a81dded211f.
Dynamically linked, not static: the table below lists every shared
library it needs, and ldd on the asset prints the same list.
| Asset | html-render_0.3.0-beta.7_linux_amd64.tar.gz |
| Binary | html-render, 25M, single file |
| Runtime deps | 13 system libraries: glibc, libssl/libcrypto, freetype, fontconfig, expat, png, zlib, bz2, brotli |
| glibc floor | 2.35 (asserted at build time) |
| Built on | ubuntu-22.04 |
| Blitz pin | see Cargo.toml at this tag |
Install:
V=v0.3.0-beta.7
curl -fsSL -O "https://github.com/leshchenko1979/html-render/releases/download/$V/html-render_0.3.0-beta.7_linux_amd64.tar.gz"
sha256sum -c <(curl -fsSL "https://github.com/leshchenko1979/html-render/releases/download/$V/checksums.txt")
tar xzf "html-render_0.3.0-beta.7_linux_amd64.tar.gz"
install -m755 html-render /usr/local/bin/html-renderNot a browser: renders JS-free HTML/CSS to PNG.
Measured on a 12-slide, 4 MB deck at 2560x1440: 5.37 s against
headless Chromium's 13.36 s (2.49x faster). The RSS figures published
for that deck — 93.4 MiB against 858 MiB — are largest single
process, not the process tree; measured on a deck in the repo,
Chromium's tree peaks 5.7-6.1x above its largest process while
html-render is one process, so the real memory gap is wider than the
9.19x those two numbers give. 4.13% of pixels differ by more than 16
levels (glyph antialiasing, not layout), and ten renders of the same
input are byte-identical. Source of these figures: GATES.md.
v0.3.0-beta.6
Blitz-based renderer CLI, built from 49506d16c8648baf4333a510982e0147a82ed506.
Dynamically linked, not static: the table below lists every shared
library it needs, and ldd on the asset prints the same list.
| Asset | html-render_0.3.0-beta.6_linux_amd64.tar.gz |
| Binary | html-render, 25M, single file |
| Runtime deps | 13 system libraries: glibc, libssl/libcrypto, freetype, fontconfig, expat, png, zlib, bz2, brotli |
| glibc floor | 2.35 (asserted at build time) |
| Built on | ubuntu-22.04 |
| Blitz pin | see Cargo.toml at this tag |
Install:
V=v0.3.0-beta.6
curl -fsSL -O "https://github.com/leshchenko1979/html-render/releases/download/$V/html-render_0.3.0-beta.6_linux_amd64.tar.gz"
sha256sum -c <(curl -fsSL "https://github.com/leshchenko1979/html-render/releases/download/$V/checksums.txt")
tar xzf "html-render_0.3.0-beta.6_linux_amd64.tar.gz"
install -m755 html-render /usr/local/bin/html-renderNot a browser: renders JS-free HTML/CSS to PNG.
Measured on a 12-slide, 4 MB deck at 2560x1440: 5.37 s against
headless Chromium's 13.36 s (2.49x faster). The RSS figures published
for that deck — 93.4 MiB against 858 MiB — are largest single
process, not the process tree; measured on a deck in the repo,
Chromium's tree peaks 5.7-6.1x above its largest process while
html-render is one process, so the real memory gap is wider than the
9.19x those two numbers give. 4.13% of pixels differ by more than 16
levels (glyph antialiasing, not layout), and ten renders of the same
input are byte-identical. Source of these figures: GATES.md.
v0.3.0-beta.5
Blitz-based renderer CLI, built from a50e65eaca73be3497680ba9d5ede54dc7856be9.
Dynamically linked, not static: the table below lists every shared
library it needs, and ldd on the asset prints the same list.
| Asset | html-render_0.3.0-beta.5_linux_amd64.tar.gz |
| Binary | html-render, 25M, single file |
| Runtime deps | 13 system libraries: glibc, libssl/libcrypto, freetype, fontconfig, expat, png, zlib, bz2, brotli |
| glibc floor | 2.35 (asserted at build time) |
| Built on | ubuntu-22.04 |
| Blitz pin | see Cargo.toml at this tag |
Install:
V=v0.3.0-beta.5
curl -fsSL -O "https://github.com/leshchenko1979/html-render/releases/download/$V/html-render_0.3.0-beta.5_linux_amd64.tar.gz"
sha256sum -c <(curl -fsSL "https://github.com/leshchenko1979/html-render/releases/download/$V/checksums.txt")
tar xzf "html-render_0.3.0-beta.5_linux_amd64.tar.gz"
install -m755 html-render /usr/local/bin/html-renderNot a browser: renders JS-free HTML/CSS to PNG.
Measured on a 12-slide, 4 MB deck at 2560x1440: 5.37 s against
headless Chromium's 13.36 s (2.49x faster). The RSS figures published
for that deck — 93.4 MiB against 858 MiB — are largest single
process, not the process tree; measured on a deck in the repo,
Chromium's tree peaks 5.7-6.1x above its largest process while
html-render is one process, so the real memory gap is wider than the
9.19x those two numbers give. 4.13% of pixels differ by more than 16
levels (glyph antialiasing, not layout), and ten renders of the same
input are byte-identical. Source of these figures: GATES.md.
v0.3.0-beta.4-r2
Warning
YANKED — do not pin this release. Encodes a rebuild in the version string: under SemVer §11 0.3.0-beta.4-r2 sorts above 0.3.0-beta.10, so a resolver reads this eleven-day-old release as the newest.
Superseded by v0.4.0, which is the first release under the current version scheme. The asset below is left in place because published artefacts are immutable.
Blitz-based renderer CLI, built from e730eac6fdd3a429abd3cb7841e0d5c83289e88e.
Dynamically linked, not static: the Runtime deps row below lists every shared
library it needs, and ldd on the asset prints the same list.
| Asset | html-render_0.3.0-beta.4-r2_linux_amd64.tar.gz |
| Binary | html-render, 25M, single file |
| Runtime deps | 13 system libraries: glibc, libssl/libcrypto, freetype, fontconfig, expat, png, zlib, bz2, brotli |
| glibc floor | 2.35 (asserted at build time) |
| Built on | ubuntu-22.04 |
| Blitz pin | see Cargo.toml at this tag |
Install:
V=v0.3.0-beta.4-r2
curl -fsSL -O "https://github.com/leshchenko1979/html-render/releases/download/$V/html-render_0.3.0-beta.4-r2_linux_amd64.tar.gz"
sha256sum -c <(curl -fsSL "https://github.com/leshchenko1979/html-render/releases/download/$V/checksums.txt")
tar xzf "html-render_0.3.0-beta.4-r2_linux_amd64.tar.gz"
install -m755 html-render /usr/local/bin/html-renderNot a browser: renders JS-free HTML/CSS to PNG.
Measured on a 12-slide, 4 MB deck at 2560x1440, in --scroll mode: 5.37 s
against headless Chromium's 13.36 s — 2.49x faster. Mode matters: the default
--clip mode on the same deck takes 5.72 s, so the headline is the fastest mode,
not the one you get without asking. The RSS figures published for that deck —
93.4 MiB against 858 MiB — are largest single process, not the process tree;
measured on a deck in the repo, Chromium's tree peaks 5.7-6.1x above its
largest process while html-render is one process, so the real memory gap is
wider than the 9.19x those two numbers give. 4.13% of pixels differ by more than
16 levels (glyph antialiasing, not layout), and ten renders of the same input are
byte-identical for documents whose resources are local or data: URIs.
Every figure is reproduced from your own deck with
python3 tools/gates.py --deck <path>; see GATES.md at this tag.