Mutation testing found behaviour no test held, now covered:
- a Group 4 halftone above 450 ppi isn't resampled and stays byte for
byte, alone by default and behind ASCIIHexDecode with Retain;
- under Retain a resampled Flate scan stays Flate, though Group 4 would
be smaller (ragged bands);
- sixteen-bit grey of only 0x0000 and 0xFFFF keeps its bytes;
- a busy Group 4 image behind ASCIIHexDecode, whose code is over three
times its raster, decodes within the fax bound and is stored again as
Group 4 alone. Redaction's busy-fax test is that busy again: since the
bound for other images is three times the raster, the old checkerboard
fitted it, and dropping the fax bound passed.
The redaction README says three times, and doc_optimize's description
says what Retain keeps: CCITT stays CCITT, anything else becomes Flate.
From reviews with the pdfcraft-review skill.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P6yUtvWGPsJpMLyh6pxSXG
- The optimize README, the crate doc and both parity notes say a
halftone or dither, and an image wider than a fax row, isn't averaged
down.
- The README's "Left alone" list is the colour and grey settings'; image
masks, decode arrays and CCITT go to Monochrome Images now. Its
monochrome "Not yet" names an image's own /SMask or /Mask image again
(only images drawn on a page are visited), as the parity note does;
the parity note's row limit is Group 4's.
- docs/architecture.md says CCITT decodes through pdfcraft-filters for
export, redaction and the optimizer, not only the renderer.
- docs/target-app-parity.md no longer lists monochrome downsampling as
missing.
From reviews with the pdfcraft-review skill.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P6yUtvWGPsJpMLyh6pxSXG
- Halftones and dithers: most of their new pixels cover both black and
white, and taking the sample that covers more turned their tones
solid (a 50% screen black, a 25% one white). An image whose new pixels
mostly cover both is now left at its resolution; text covers both only
along its edges and is still averaged down. downsample_bits says how
many new pixels cover both.
- A counter for each column: an image wider than a fax row
(MAX_CCITT_COLUMNS) isn't averaged, so a strip of 80 million columns
can't make the counters 640 MB.
- A fax image's decoding is bounded at about a byte a pixel, as
redaction's is (decoded_image_within), not 1 GiB; behind a Flate bomb
it is left alone.
- An indexed image's tie went to sample 0, which an inverting decode
array maps to entry 1, though the docs say entry 0. It goes to the
sample that decodes to entry 0 now (ISO 32000-2 §8.9.5.2).
Each has a test that fails without it.
From reviews with the pdfcraft-review skill.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P6yUtvWGPsJpMLyh6pxSXG
doc_reduce, the engine's reduced_bytes doc, the optimize README's
presets and the optimize.reduce parity note still said every image
above 225 ppi goes to 150 ppi; black-and-white images go to 300 ppi
above 450 ppi, as CCITT Group 4 when smaller.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Create PDF and the optimizer now store black-and-white scans as Group 4,
and redaction couldn't decode CCITT, so redacting one word on such a
scan removed the whole page image. A CCITT image (the last filter,
after any others) now decodes through Document::decoded_image and is
cleared in a Flate copy like any other image; one that can't be decoded
in full is still removed whole.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The README said an indexed image's palette decides which entry is the
dark one on a tie, but dark_is_one never reads the palette: entry 0
wins whatever its colour. The README, the function's comment and the
parity note now say so (from FredApps' review).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A 2 × 2 tie always gave sample 0, black in grey and the painted part of
an image mask with the default decode array. With an inverting one,
[1 0], 0 is white or unpainted, and in a separation 0 is no ink, so a
line half a new pixel wide could vanish. dark_is_one says which sample
is dark from the colour space and decode array: 0 in grey and of a
mask, 1 in a separation or a DeviceN of one colourant, each the other
way round when the decode array inverts; an indexed image keeps entry
0, as its palette, not the decode array, says which is darker.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BfMGHQNgdq5yf5tKZHGR3z
The PDF Optimizer and Reduce File Size downsampled colour and grey images
but not black-and-white ones. Settings::mono is now MonoSettings {
downsample, target_ppi, above_ppi, compression }, 300 ppi above 450 ppi
and CCITT Group 4 by default (Monochrome Images ▸ Downsample to;
doc_optimize mono: { downsample, ppi, above_ppi, compression }).
A 1-bit image or image mask (no filter, general-purpose filters, or CCITT
fax after any of them) and eight-bit grey of only black and white are averaged by
area: each new pixel takes the sample covering more of it, a tie 0, so a
dark line half a new pixel wide stays. They are stored as Group 4, or as
1-bit Flate when that is smaller; with Retain a resampled image keeps its
kind of compression. Images with a soft mask keep their size.
CCITT decoding moves from Export ▸ Images into pdfcraft-filters
(decode_ccitt, CcittParams, through hayro-ccitt), and
pdfcraft_cos::Document::decoded_image reads a fax image's /DecodeParms
for it (undoing filters before CCITT, such as ASCII85Decode, with any
/DecodeParms given by reference made direct first), so export
and the optimizer share one reader, and Group 4 scans can be
downsampled. A 600 ppi text page went from 79 KB to 40 KB.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BfMGHQNgdq5yf5tKZHGR3z
- A noisy 640 × 800 soft mask under 1200 × 1500 text: resampled to the
image's new size it shrinks only a little, by less than the grey
settings' JPEG adds over the packed image, so packing wins; not
counting the new mask passed every test before.
- The hex-text half of the bomb test also stores eight-bit
black-and-white, which must pack, as its comment says.
- The optimize README, the mono-compression parity note and both
progress logs say the soft mask counts.
- The Copy Image parity note no longer lists CCITT as missing: it
copies through Export's decoder, which reads CCITT now.
From reviews with the pdfcraft-review skill.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P6yUtvWGPsJpMLyh6pxSXG
- Eight-bit black-and-white text under a noisy soft mask: the grey
settings win, as they shrink the mask far more than packing saves.
Counting the packed image without its mask passed every test before.
- The packing test checks the report's byte totals exactly.
- Export reads Group 3 with end-of-line codes, with rows on a byte
(EncodedByteAlign), with both, and mixed (/K 1, each row tagged
one-dimensional) with end-of-line codes and without. Decoding /K 1 as
one-dimensional, or ignoring EncodedByteAlign, now fails a test.
/EndOfLine is passed to the decoder, but no input here tells it apart:
the decoder accepts end-of-line codes either way, as ISO 32000-2
(Table 11) requires.
From reviews with the pdfcraft-review skill.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P6yUtvWGPsJpMLyh6pxSXG
An eight-bit grey image of only black and white was packed to one bit
and stored, and the grey settings never saw it. For a dithered scan that
is bigger than what Reduce File Size made of it before, downsampled to
JPEG: 1-bit Flate can't shrink the dots (a 1200 × 1500 dither at 300 ppi
was 157 KB packed, 67 KB as before). Both are made now, and the smaller
as stored wins, counting a soft mask resampled with the image. Clean
text still packs to one bit. The packing also decodes no further than
its samples can need, as the other image steps do now.
From reviews with the pdfcraft-review skill.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P6yUtvWGPsJpMLyh6pxSXG
Packing eight-bit black-and-white grey to one bit decoded with
Stream::decoded, which ignores /DecodeParms given by reference. With PNG
predictor 10 each row then kept its tag byte, a 0, so the shifted rows
still passed as black and white and were packed and stored skewed. It
now decodes through Document::with_direct_parms (#4304), and the test
of images with their parameters by reference covers this case.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BfMGHQNgdq5yf5tKZHGR3z
A CCITT image whose /K, /Columns or /Rows is an indirect integer was read
with the defaults (K 0, 1728 columns), and one whose /Rows is below its
/Height was cut short and skipped. The numbers are now resolved, and the
image is decoded up to its /Height rows whatever /Rows says, still bounded
by the size the export checked (review of #4335).
The doc comment of the optimizer's bilevel step now says what Retain
leaves to the grey settings.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BfMGHQNgdq5yf5tKZHGR3z
Group 4 images (now written by Create PDF and the optimizer) would
otherwise stop exporting: they were reported as "can't be exported
yet". A CCITT fax image (its only filter) is decoded through
hayro-ccitt with its /DecodeParms (K, Columns, Rows, EndOfBlock,
EndOfLine, EncodedByteAlign, BlackIs1), masks included; /Columns has to
be the image's width, so a hostile value can't size the buffers.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BfMGHQNgdq5yf5tKZHGR3z
An eight-bit grey image whose samples are all 0 or 255 is packed to one
bit per pixel and then stored as Monochrome Images asks (Group 4, or
JBIG2 with #4320), when that is smaller as stored; its decode array
stays. Indexed images, a colour-key /Mask and Retain are left alone.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BfMGHQNgdq5yf5tKZHGR3z
- Not only Export ▸ Images refuses a CCITT image until #4335: so does
everything that exports an image through it. Save Image As refuses
it, Copy Image fails, and Word, PowerPoint and HTML export leave the
picture out, with the page not reported. The optimize README and the
mono-compression parity note name them.
- optimize.mono-compression is partial now, so docs/target-app-parity.md
counts it on the table's own measurement: K Optimize and standards 15
shipped and 11 partial of 58 (35.3%), breadth 57.0% (423 + 75 of 808),
mainstream 63.3% (292 + 45 of 497), with a status line and a revision
row. ROADMAP.md's breadth rows and its K row say the same.
From reviews with the pdfcraft-review skill.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P6yUtvWGPsJpMLyh6pxSXG
The optimizer's images decode no further than twice their samples plus
a byte a row, meant for ASCIIHex text under Flate; but hex text broken
into lines, as it usually is, is a little more than twice, so such an
image was left alone as if it were a decompression bomb (a 300 dpi
letter scan stayed Flate). It is three bytes for each now, a separator
for every two digits, and a test stores both images as hex text in
lines of 64 digits.
From reviews with the pdfcraft-review skill.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P6yUtvWGPsJpMLyh6pxSXG
- Redaction doesn't decode CCITT, so a mark over part of an image the
optimizer stored as Group 4 removes the whole image (nothing leaks),
and Export ▸ Images refuses it. The optimize README and the
mono-compression parity note say so. Later PRs of this stack add both
(#4335 export, #4384 redaction).
- doc_reduce's description, Engine::reduced_bytes and docs/cli-batch.md
now list the black-and-white step; docs/target-app-parity.md's
optimize row gives the measured counts and no longer lists CCITT as
missing; the M11 milestone row lists it as done.
From reviews with the pdfcraft-review skill.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P6yUtvWGPsJpMLyh6pxSXG
A test now covers what qualifies and what doesn't: an ICC profile of
one component and a DeviceN of one colourant convert, three components
and two colourants (more than a bit a pixel) don't; an image mask
converts with /BitsPerComponent 1 given, not with 8; an image of exactly
80 million pixels converts, one row more doesn't. The byte totals are
checked exactly. Each of those rules could be broken before with no
test failing (cargo mutants). one_component's comment now says /G
counts with DeviceGray, as the code does.
From reviews with the pdfcraft-review skill.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P6yUtvWGPsJpMLyh6pxSXG
A 1-bit image was decoded for Group 4 up to MAX_DECODED (1 GiB), so a
few kilobytes of Flate could cost a gigabyte, and the same held for the
samples of a colour or grey image read to resample it (pixels). Both now
stop at twice the samples plus a byte a row and 4 KiB, as redaction
does; data past that can only be a decompression bomb, and the image is
left alone.
From reviews with the pdfcraft-review skill.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P6yUtvWGPsJpMLyh6pxSXG
Document::with_direct_parms resolved an indirect /DecodeParms and its
array entries but not the numbers inside (/Predictor 9 0 R), which
Dict::int doesn't follow. A 1-bit image with such a predictor was
decoded without it, its rows shifted by their tag bytes, and Group 4 of
that stored in its place; resampling read images the same way. It now
keeps, of each parameters dictionary, the numbers and flags the decoders
read, resolved, and /JBIG2Globals as a reference. Only those small
values are copied (an entry that is no dictionary is null), so a file
naming one big object from every key and filter can't have it copied a
hundred times.
The change is #4384's, moved down so that every PR stacked on this one
reads such parameters.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P6yUtvWGPsJpMLyh6pxSXG
Stream::decoded reads only direct parameter dictionaries, so an image
whose /DecodeParms is a reference (whole, or as array entries) was
decoded without its predictor. For a 1-bit image with a PNG predictor
the rows came out shifted by their tag bytes, Group 4 of that was still
smaller, and the picture was replaced by a skewed one; resampling and
merging image strips read the same way.
pdfcraft_cos::Document::with_direct_parms gives the stream with the
parameters its filters use made direct: only the entries at the
filters' places, and none for a chain longer than any decoder takes,
so at most MAX_CHAIN objects are copied. The optimizer's pixels() and
the Group 4 step decode through it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BfMGHQNgdq5yf5tKZHGR3z
A bare /Indexed, /I or /CalGray is no colour space on its own, so it can
only name a resource, which might be RGB: such an image is now left
alone instead of being read as one bit per pixel. doc_optimize refuses a
mono that isn't an object or a compression that isn't a known name,
instead of reading it as the default (from CodeRabbit's review).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BfMGHQNgdq5yf5tKZHGR3z
pdfcraft_filters::encode_g4 encodes packed 1-bit rows as Group 4 (ITU-T
T.6, /K -1) through the fax crate (MIT, already in the tree via tiff),
with rows capped at 32,767 samples so its 16-bit positions can't
overflow; round trips are checked against hayro-ccitt.
The optimizer (Settings::mono, CCITT G4 by default, so Reduce File Size
too) re-encodes a 1-bit image of one colour component, or an image
mask, stored with no filter or general-purpose ones, when that is
smaller as stored. The samples decode to the same bits, so colour
space, decode array and masks stay. PDF Optimizer gets a Monochrome
Images row and doc_optimize a mono option. optimize.mono-compression
goes from planned to partial (JBIG2 and monochrome downsampling not
yet).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BfMGHQNgdq5yf5tKZHGR3z
Clean Up removes named destinations that nothing in the document refers
to. referenced_names counted the /D of GoTo, GoToR and GoToE actions
alike. The /D of a GoToR or GoToE action names a destination in another
document, so a local destination with the same name was kept although
nothing in the document used it.
Only a GoTo's /D now counts as a use. A /Dest key and a GoTo's /D are
read as before.
Test: links with a GoTo, a GoToR and a GoToE action name three
destinations; optimize() keeps the one the GoTo uses and removes the
other two.
ROADMAP.md: kept both progress lines.
crates/optimize/README.md: PR's decoder sentence plus the target-version section from main.
Cargo.lock: auto-merged.
Conflicts: ATTRIBUTION.toml: took main's side (hashes recomputed after the merges). crates/optimize/src/lib.rs, crates/automation/src/lib.rs, tools.rs: Settings and tool schema keep object_streams (#3227) and lzw_to_flate (#4181); doc_optimize description names both options. crates/optimize/tests/optimize.rs: both sides' imports and tests kept.
Conflicts: docs/roadmap.md: all status lines kept, M10 row from main, M11 row with #3227's two sentences. parity/acrobat-features.toml: save-with-object-streams entry from #3227 with its test paths under tests/it/.
Reduce File Size removes named destinations that nothing refers to. It
collected the referenced names by reading /Dest and a GoTo's /D
directly, so when a link's /Dest was an indirect reference to the name
string (valid PDF) rather than the string itself, the name was not
counted. The destination it used was then treated as unreferenced and
deleted, silently breaking the link.
referenced_names now resolves /Dest and /D before reading the name, so
an indirectly referenced destination is kept.
Test: a link whose /Dest is a reference to the name string keeps its
destination through optimize(); an unused destination is still removed.
Reported-by: BS.ehlas.cz
Optimize reads a JPEG's header again for every group of fragments it tries,
and one image can be drawn many times. The scan skipped stray bytes between
segments one at a time, so a JPEG with megabytes of them after SOI cost a
full walk on every call: 64 references to one such image delayed the first
progress call by seconds.
The scan now treats a stray byte as an unreadable header, which leaves the
image as it is, and stops after 4096 markers and fill bytes. Each step jumps
over a whole segment, so the work no longer grows with the stream's size.
Redaction reads page content and form XObjects strictly (upstream) and
within one page's decode budget (this branch): both now go through
DecodeBudget::decode_strict.
A run of images that is not one picture was searched again from each of
its groups, collecting the rest of the run every time, so many images drawn
over one another made a page quadratic. The search reads at most 16 groups
per content operator (at least 65 536) and leaves the rest of the page as
it is once that is spent.
The pixel reader capped the image size the dictionary gives, then decoded
the JPEG at the size its header gives and compared afterwards, so a small
image dictionary over a huge JPEG allocated the huge one first. The reader
takes the size from the JPEG header and refuses a mismatch before decoding.
The fragment merge multiplied the merged width by a height cast from a
float without checking either fit. A tiny fragment rectangle gives a huge
height: the multiplication overflowed (a panic in debug builds, a wrap past
the pixel limit in release builds) and the u32 cast saturated, sizing an
enormous allocation. The height must now fit u32 before the pixel limit is
checked, and the buffer size uses checked multiplication.
Image measurement and fragment merging decoded every /Contents entry
and every form occurrence with a fresh 1 GiB allowance, so repeated
references to one compressed stream multiplied memory. Both charge a
per-page DecodeBudget. A page whose content does not decode within it
is left unmerged, and its images are not resampled, since a use may be
missed; fragment merging no longer rewrites a page from content it
could not fully decode.
ROADMAP.md: progress-log line added at the top
docs/roadmap.md: the M11 note applied onto main's row (which #3622 extended)
crates/optimize/tests/optimize.rs: new tests appended after main's
Optimize decoded DCT images with the generic JPEG decoder, which converts
three components from YCbCr to RGB whenever the JPEG itself doesn't say
otherwise, and then dropped /DecodeParms. An RGB JPEG stored untransformed
(/ColorTransform 0 and no Adobe marker) therefore came out in other colours
when recompressed, resampled or merged from strips.
The samples are now read with the colour transform ISO 32000-2 §7.4.8
gives them: an Adobe APP14 marker's flag, else /ColorTransform (default 1).
Untransformed samples are taken as stored, and the result is written as a
plain YCbCr JPEG (or Flate), which needs no /DecodeParms. When the marker
and /ColorTransform disagree, a value is out of range, the parameters are
not paired with the filter as §7.4 pairs them, or the decoder reads the
components otherwise (named R, G and B), the image is left as it is.
The tests draw each image with PdfCraft's renderer before and after
Optimize: RGB JPEGs with and without an Adobe marker under each
/ColorTransform, CMYK JPEGs (still left alone) and JPEG strips merged into
one picture.
Part of #1201.
* fix(optimize): count what form fields and comments draw with by kind
The space audit claimed the AcroForm and annotations before any page
resources, so the fonts in /DR and in widget and comment appearance
streams were counted as Forms or Comments. A fillable form with an
embedded field font showed mostly "Forms" and almost no "Fonts".
Resources are now sorted by kind first, wherever they are used: page
resources, the form's default resources and every annotation
appearance (one stream or a dictionary of states). A form XObject's
other entries now count as X Object Forms instead of overhead.
Sorting nested forms' resources recursed once per form, so a long
chain of nested forms overflowed the stack. The recursion is now
bounded at 64 levels; deeper forms are still counted as X Object
Forms through the iterative walk.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(ui): line up the Space Audit columns and say what it shows
The percentages were drawn at the dialog's far edge, away from their
header and outside the striped rows: egui gives a grid's last column
the rest of the window, and the right-to-left layout used to
right-align the numbers filled all of it. Every cell is now laid out
first and given exactly its column's width; numeric columns and
their headers are right-aligned, and the description column takes
the rest, so the stripes span the dialog.
A caption under the title says the table is what takes up the file's
space now (not what optimizing would save), translated in every
catalogue. The parity notes no longer call the space audit missing.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Refresh catalog hashes after merging main
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
The effective-resolution walk skipped any form XObject already visited
once when it was nested more than four deep, with one seen-set for the
whole document. A later, larger use of the same form was never measured,
so Reduce File Size and Optimize PDF resampled the image for its smallest
use: a 1200 px image drawn 1 in and 8 in wide through six nested forms
came out 150 px wide, about 19 ppi on the large copy.
- Cycles are detected on the current path of forms, not document-wide.
- A form with its own resources drawn again at the same scale (only
moved) is skipped, as that can't change any image's resolution; forms
without resources draw their caller's images and are always walked.
- Repeat walks are capped at 100,000. If the cap or the 12-level depth
limit leaves a use unmeasured, every image gets INFINITY: recompressed
as the settings say, never resampled.
Fixes#1811
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Brandon Thomas <echelon@gmail.com>
