This changelog format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

• modify_layer_configs operation: add, update, delete, rename or promote to default the named layer configurations (/OCProperties /Configs); dump_layers now reports each configuration's locked layers and creator

• resample_images now takes mono_dpi=<n> and threshold=<x>

• subset_fonts takes keep_type1, leaving Type 1 fonts unconverted (Poppler draws a font converted to CFF with slightly different glyph edges)

• shrink operation: makes a PDF smaller in one step, at lossless (default), balanced, strong or extreme, running pdftl's size-reduction passes in a benchmarked order; target=<size> picks the least lossy settings that fit

• delete_tags operation: deletes the structure tree (tags); pages look the same, but the file is no longer accessible

• compact_content operation: rewrites content streams in their shortest equivalent form

• merge_font_subsets operation: merges the per-page subsets of one TrueType or Type1C (CFF) font into one font

• deduplicate_xobjects operation: merges identical Form XObjects and link appearances

• round_text_positions operation: writes text positions with fewer digits, moving no glyph more than a tolerance

• photos_to_jpeg operation: re-encodes losslessly stored photographs as JPEG when the JPEG keeps their pixels and local structure (needs the photos-to-jpeg extra)

• mrc_compress operation: splits scanned pages into a 1-bit text layer over low-resolution colour layers (needs the mrc-compress extra)

• recompress, compress_xmp, drop_thumbnails and deflate zopfli output options: recompress streams at maximum level, compress XMP metadata, drop page thumbnails, use zopfli

• render works in a pipeline, handing a later stage an image-only PDF

• PDFTL_PROFILE_MEMORY=<stage>|all writes a memory profile of a CLI stage, and stages using more than PDFTL_MEMORY_THRESHOLD MB say so

• Interactive password prompt support for encrypted PDFs (using PROMPT or A=PROMPT).

• Parsing an embedded Type 1 font program is now bounded in steps, operand stack and memory, so a crafted font can no longer hang pdftl or exhaust memory

• Interpreting a glyph's charstring (reading CFF widths, converting Type 1 to CFF) is now bounded in work and subroutine nesting; a glyph with runaway subroutines is skipped rather than hanging or crashing the operation

• modify_layers: a specific target now overrides a conflicting action for all, so hide all show X reliably leaves X visible (it depended on Python's hash seed) and strip all keep X keeps X; conflicting actions on the same target are an error

• dump_layers, modify_layers, modify_layer_configs and layer creation in overlay, barcode and redact no longer crash on malformed optional content: a lone dictionary where an array belongs (e.g. /Configs), junk array entries, or a non-dictionary /D or /Usage. dump_layers also accepts an /Intent given as a single name, as the spec allows

• modify_layers strip/merge now also removes the layers from /Locked, /RBGroups and the /AS auto-state arrays

• modify_layers strip/merge now handles content controlled by several layers through an optional content membership dictionary (OCMD) by its /P policy or /VE expression: such content was left untouched when marked, and an XObject was dropped (strip) or made permanent (merge) whenever any one of its layers was targeted

• modify_layers strip/merge now applies to annotations: those on a stripped layer are deleted with their popups (form fields are also removed from the form), those on a merged layer become permanent, and layered content in appearance streams is stripped or merged like page content

• render with no output writes page_1.png, page_2.png, ... as documented

• optimize_images never makes an image bigger, also optimises inverted bitonal images (without changing a pixel), and at low no longer re-encodes JPEGs

• resample_images, recolor_images, modify_images and render bound the memory they use, and running out of memory prints an error, not a traceback

• simplify_vectors leaves alone a content stream it cannot fully parse, instead of dropping the unreadable parts

• excise and redact no longer move surviving text: tidying the rewritten content stream dropped the line spacing a removed line's TD had set, moved T* ahead of a Tm, and treated a Tz restored by Q (or inherited by a Form XObject) as redundant

• excise and redact keep a colour or font set just before a Q with no matching q

• simplify_vectors no longer makes a content stream bigger (it keeps the original when the simplified one would not compress smaller), uses less memory, and skips content streams over the new max_stream_size (default 16MB) instead of exhausting memory on them

• subset_fonts no longer changes how text renders: it keeps a TrueType or CFF font's original font-wide bounding box (a recomputed, tighter one altered Poppler's rendering)

• subset_fonts keeps only the glyphs a PDF draws, and drops font tables PDF renderers never read (Office documents' fonts come out about half the size)

• subset_fonts no longer gives a simple TrueType font widths keyed by Unicode (/LastChar 8217), which made MuPDF draw each line's text on top of itself

• deduplicate_images and deduplicate_xobjects now merge copies that differ only in the obsolete /Name key, which ImageMagick sets per copy (a 36 MB photo PDF drops to 12 MB)

• The oxipng step of save-time recompression now works on Windows, where oxipng could not open the temporary file

• uncompress now really stores streams uncompressed: streams that were already Flate-compressed in the input used to be copied still compressed

• Images decoded through the low-level fallback (used when pikepdf cannot extract an image, for example by recolor_images and export_images) now have their /Decode array applied, like every other image, instead of coming out inverted; encoded image data that happened to be longer than the raw pixel size is no longer misread as raw pixels

• resample_images and modify_images skip images Pillow cannot represent (such as 5-ink DeviceN "hi-fi" images) instead of aborting the whole operation

• Embedded Type 1 fonts with no eexec trailer (/Length3 0, as TeX-produced PDFs commonly have) now get their widths read and patched, rather than silently skipped

• Embedded CFF fonts whose Top DICT omits charset (meaning the standard ISOAdobe charset) now get their widths read and patched; patching such a font, or a CID-keyed one missing its charset, no longer crashes

• resample_images, modify_images, import_images and recolor_images no longer invert images with a non-default /Decode array (such as [1 0] bitonal scans, which came out as negatives)

• Over-aggressive content-stream cleanup in excise and redaction fixed

• unpause: raise error on unknown arguments instead of silently ignoring them

• booklet and montage: validate and reject unknown configuration keys

• set_info: allow setting PdfID0 metadata on PDFs lacking an existing trailer /ID array

• delete_attachments: annotations of removed attachments are now removed too; minbytes/maxbytes compare the file size, not the compressed stream length

• dump_tags: linter now checks every inline (direct) structure element, not just the first

• simplify_vectors, excise: stroked rectangles no longer lose their fourth side; stroke merging no longer drops the closing edge of closed subpaths

• set_info: an invalid PageLabelStart now skips the label, as its warning says, instead of writing /St anyway

• dump_bookmarks, dump_data, toc and delete_bookmarks: destinations that target structure elements now resolve to the correct page when the page reference is on a later inline child

• stamp_fields: inline widget kids after the first are no longer skipped

• Documentation fixes:

  • Correct FORCE_COLORS to FORCE_COLOR
  • Correct input handle lifetime documentation
  • Clarify handle scoping for page ranges in cat (e.g., B2,4 vs B2,B4)

• barcode: correct orientation on rotated pages

• delete_images, export_images, import_images, recolor_images, and resample_images now target inline images (BI/ID/EI content stream blocks), not just XObject images

• dump_annots and delete_annots now support filtering by nested annotation properties using slash-separated key paths, both by value and existence

• New output options:

  • drop_meta: discard all metadata
  • drop_vendor_extensions: discard private vendor/application data hung directly off the document catalog
  • drop_xmp_streams: discard all XMP metadata streams

• usage now reports private vendor/application data hung directly off the document catalog in a new vendor_extension category, rather than as part of document_structure

• modify_images: despeckle transformation now takes a size parameter

• optimize_images accepts jbig2_group_size parameter to allow better compression

• help now prioritizes nearby keywords/help topics in the same pipline stage rather than always choosing the first one

• dump_annots and delete_annots handle malformed arguments better

• redact no longer leaves single-character regex matches (e.g. /e/) in the content stream; glyph deletion now uses real overlap/containment against the glyph's own box (excise's new glyph_overlap option) instead of only a center-point test

• modify_images now finds images drawn through Form XObjects (common in scanned PDFs) instead of reporting that no images were found

• optimize_images no longer creates an images directory next to the output file (or writes into an existing one); intermediate files now go in a temporary directory that is removed afterwards

• optimize_images: the jpg_quality=<n> option is no longer silently ignored

• dump_images and modify_images no longer crash on images whose /Filter is an empty array

• server: oversized uploads are now rejected with a properly delivered 413 on Windows, instead of a connection error.

• server: correct stale file logging in verbose mode

• dump_bookmarks: friendlier launch, goto_remote, named_action keys for /Launch, /GoToR, and /Named outline actions (previously a raw ISO action dict)

• excise outputs cleaner content streams

• dump_images PPI calculation fixed to take CTM into account

• dump_images now shows image masks

• dump_bookmarks/update_bookmarks: fix a crash on outline items with both /Dest and /A (keeps /Dest)

• Bump minimum pikepdf version to fix AttributeError when reading page rotation during cat operations.

• prune_resources output option

• deduplicate_icc_profiles operation

• deduplicate_fonts operation

• deduplicate_images operation

• add_marks operation

• usage operation

• link_urls operation

• Can now give chop an overlap

• redact: attempt to verify redaction success when verify=true is passed

• Custom CA certificate support added to dump_signatures via new trust_roots option, for deterministic trust chain verification.

• Friendly error handling when input and output filenames agree

• redact now redacts entire matches by default. Capturing groups (...) no longer narrow the target unless redact*-named groups are explicitly used (?P<redact_foo>...).

• A warning is issued when saving will invalidate a signature

• excise and redact now remove annotations whose rectangle cannot be determined

• replace, style_text, mutate_content, simplify_vectors, recolor_vectors, dump_streams and import_streams now also recurse into tiling Patterns, ExtGState soft-mask groups, and annotation appearance streams (previously only Form XObjects, or Forms+Patterns for the vector operations). Content painted through these mechanisms was silently skipped before; it is now processed like any other reachable content stream.

• Resolved an issue where signing PDFs inside an active async event loop could fail.

• redact operation

• excise operation

• add image format selection to modify_images

• add automatic bookmark detection to update_bookmarks

• subset_fonts: add subset tags to font names

• add_text: improve system font matching

• modify_images: improved posterize to make colors more faithful

• dump_text: correctly handle type 3 fonts without /ToUnicode where possible

• subset_fonts: improve Type 1 conversion performance

• style_text: use stroke=none to remove text stroke

• subset_fonts operation

• style_text: adding a stroke, by default, now uses the text fill color (if any) as the stroke color

• render: API now supports explicit format=png|jpg|jpeg|pdf output selection. Image formats return a ZIP archive; format=pdf returns a single rasterized PDF. The CLI continues to determine output format from output.

• server:

  • UI enhancements to builder interface
  • server pipeline now takes understands @A per-operation arguments, allowing use of stamp, for example

• dump_fonts and export_fonts now correctly scan annotation appearances for fonts

• unpack_files now correctly returns a zip archive when called via the server

• stamp, multistamp, background, multibackground: a PDF file argument to use as overlay/underlay is now required, per the documentation

• optimize_images: better errors on missing ocrmypdf system dependencies

• delete, delete_blank, unpause, insert, move: page labels (/PageLabels) are now preserved and remapped across page deletion, insertion, and reordering, so surviving pages keep their original logical page numbers (e.g. roman-numeral front matter) instead of losing custom numbering.

• burst: each output chunk now preserves the ORIGINAL document's page labels/numbering rather than restarting at "1" — e.g. splitting a book by chapter keeps each chapter's pages numbered as they were in the whole book. This is an intentional deviation from classic pdftk burst behaviour, which has no equivalent concept.

• dump_bookmarks: strip trailing nulls from outline titles and metadata, like most PDF readers

• server:
  • Replaced a racy multiprocessing.Queue-based IPC mechanism for subprocess-isolated operation execution with a Pipe-based protocol that correctly handles worker timeouts, crashes, and cleanup. This fixes intermittent hangs, leaked worker processes/pipes, and dropped/incorrect exception types under concurrent load.
  • Fixed a timeout error message that silently dropped the configured timeout value.
  • Fixed a keep-alive connection handling bug where rejecting an oversized upload could leave the request thread blocked waiting on the remainder of an in-flight upload body.
  • Fixed a per-worker memory limit that crashed workers outright on some platforms (notably macOS) instead of degrading gracefully.

• server: bounded the number of concurrent subprocess-isolated operation workers (default: capped to available CPU count, configurable via PDFTL_MAX_CONCURRENT_WORKERS). Previously, a burst of concurrent requests could spawn an unbounded number of worker processes.

• server operation

• deskew operation

• stamp_fields operation

• embed_fonts operation

• --args <file.yml> argument insertion

• coalesce_strokes option added to simplify_vectors

• sign_field no longer changes field appearances

• style_text now adapts Tr (text rendering mode) better to user parameters

• improved project version-finding heuristics

• dump_text: if output is passed directly, we now write the text, not the PDF.

• render now initializes forms, so that they are properly rendered

• dump_signatures now correctly reports pyhanko's valid/invalid status

• import_fonts operation

• export_fonts operation

• added upscale and adaptive_threshold to modify_images

• import_images operation

• export_images operation

• import_streams operation

• add_bookmarks: new options for bookmark specs: uri, launch, named, dest, color, bold, italic

• dump_colorspaces: CalGray, CalRGB, and Lab colour spaces now resolved and reported (previously silently fell through)

• dump_tags: non-standard structure tags now reported as warnings (ISO 32000-2 §14.8)

• dump_tags: invalid structure attributes now reported as warnings per standard owner definitions (ISO 32000-2 §14.8.5)

• fill_form: multi-select choice fields now supported (ISO 32000-2 §12.7.5.4)

• fill_form: radio buttons, checkboxes, list-boxes, and nested fields now handled correctly

• generate_fdf: optional status string in generated FDF output (ISO 32000-2 Table 246)

• generate_fdf: multi-select choice field arrays now serialized correctly in FDF output

• set operation: trapped field for document trapping status (True, False, or Unknown)

• set operation: author now accepts comma-separated values for multiple authors

• Bookmarks: non-GoTo actions (e.g. Launch, Named, JavaScript) are now preserved when copying or round-tripping outlines via dump_bookmarks / add_bookmarks

• Improved verbose pipeline logging feedback loops

• Suppressed noisy metadata field conversion logging strings in dump_data

• dump_streams: output header format changed from Page <N> / Page <N> / XObject <name> (<obj>:<gen>) to Page <N> / Contents / full nested breadcrumb paths (e.g. Page 1 / XObject /Fm1 / XObject /Fm0); obj:gen suffixes are no longer shown. Form XObjects reachable via more than one path now emit a % ALIAS OF: <canonical path> stub on repeat references instead of being silently skipped

• dump_streams: content stream lines beginning with === or \ are now escaped so they don't get misread as structural headers when round-tripped through import_streams

• Improve handling of OS errors while saving to disk

• Improve handling of invalid or missing arguments in several operations

• Destination coordinates for all ISO 32000-2 explicit destination types (/FitH, /FitV, /FitR, /FitBH, /FitBV, /Fit, /FitB) now correctly transformed when rotating or scaling pages; previously only /XYZ coordinates were handled

• Null coordinate values in /XYZ destinations now correctly preserved during rotation and scaling

• fill_form: digital signature fields are now safely skipped instead of corrupted

• Bookmarks: non-GoTo outline actions (e.g. URI links, launch actions) were silently dropped when merging PDFs; they are now preserved

• Named destinations in legacy PDF 1.1 /Root/Dests dictionaries are now resolved correctly

• Structure destinations (ISO 32000-2 §12.3.2.3) in outlines now resolve to the correct page, with a fallback to page 1 if no content is found

• set operation: PDF 2.0 documents now correctly omit deprecated metadata fields from the /Info dictionary, per ISO 32000-2 §14.3.3

• delete_actions operation

• dump_actions operation

• dump_tags operation

• tag operation

• dump_streams operation

• simplify_vectors operation

• png_compression setting for the render operation

• operations are now linked in the HTML docs

• replace now recurses into XObjects by default

• delete_images global mode fixed to delete per-page images too

• burst now keeps outlines and annotations, like cat

• Prevent BrokenPipeError leaking out

• add_images operation

• diff_text operation

• dump_tables operation

• modify_images operation

• System font name and file path resolution for add_text

• block repetition of page specs via copy<N>

• multithreading for resample_images and recolor_images

• page specs now tolerate whitespace (by stripping it before parsing)

• dump_bookmarks: named destinations now resolved to page/view by default; pass no_resolve to skip

• update_bookmarks: strict schema validation; accepts structured data directly via API (bookmarks= kwarg)

• Preserve outline state (open/closed, formatting) when copying outlines

• Fix a crash when a subpipeline is passed as an operation argument

• bash completion: fix mid-word completion behaviour, update docs

• fix style_text bugs: (1) coalesce content streams (2) handle SC etc

• add pattern support to recolor_vectors

• auto-paged help output on linux/mac

• render should no longer segfault when an unknown output file extension is passed

• dump_annots now properly passing the PDF down the pipeline

• Top margin percentage bug fixed in clip and crop

• pipeline info added to verbose mode

• make parentheses optional for crop, clip, delete_blank, delete_images, and style_text, in many cases

• repetitions in page specs viarep<N>

• barcode operation

• recolor_images operation

• recolor_vectors operation

• style_text operation

• Guess shell for --completion if it is not specified

• add {count} as an alias for {n} in add_text and add_bookmarks

• unpause operation

• dump_colorspaces operation

• delete_attachments operation

• dump_fonts operation

• create operation

• resample_images operation

• grep operation

• Performance improvements in link/outline remapping

• dump_files output format changed to JSON, with more detail

• Several JSON dump output formats changed from flat lists to {"name": [...]} format

• Fix crash on outline items without a /Title

• Filter annotations in dump_annots and delete_annots

• add EACH ... JOB, documented under pdftl help pipeline

• delete_bookmarks operation

• add_bookmarks operation

• add {n} template to add_text (and add_bookmarks)

• Improve error message when too many files are opened

• When --debug is passed, optimize_images now outputs ocrmypdf debug messages

• Improve error message for invalid page specs

• dump_annots: fix crash caused by named destinations which are not valid python dictionary keys

• Updated API type stubs

• Improve broken pipe handling: suppress tracebacks when piping to a pager

• Improve test robustness (fixes #25)

• zoom operation

• fast output option

• pipx-specific installation instructions for missing dependencies

• help output uses executable name correctly

• update_bookmarks can now read bookmark data from stdin

• add_text supports markdown-style hyperlinks

• set handles common metadata, in both Info and XMP

• stamp, multistamp, background, multibackground accept page specs

• help: link to readthedocs page

• modify_layers now hooks up print/view events (/AS dictionary)

• dump_layers reports /AS dictionary hookup as 'active' state

• modify_layers operation

• dump_images operation

• highlight operation

• stamp, background, multistamp, multibackground: option to add content in a new layer

• page ranges accept portrait and landscape qualifiers

• plain render arguments are now page specs, and we can output a multi-page render PDF

• delete_blank operation

• delete_images operation

• set operation

• improve handling of unknown command line flags

• suggest possible bug reporting on errors

• CONTRIBUTING.md added

• work around a pyHanko bug affecting dump_signatures

• dump_encryption operation

• dump_bookmarks operation

• update_bookmarks operation

• montage: removed fit parameter

• pdftk compatibility: default to all permissions denied

• montage and booklet bug handling pages with negative rotation fixed

• place and add_text rotation bugs fixed and visual tests added

• crop rotation bugs fixed and visual tests added

• page ranges accept step<n> to step through a range in regular intervals

• burst accepts page specs or bookmark levels for chunk split points, and chunk size limits

• Completions updated

• dump_text now works in a pipeline, instead of crashing

• montage and booklet now handle page rotation better

• pdftk compatibility:

  • burst now outputs doc_data.txt
  • dump_data omits empty fields

• Minor shell completion performance improvements

• Improved memory management

• Fixed chop, place, and booklet operations to correctly calculate physical boundaries on pages with non-zero origins or non-standard /Rotate flags

• Paper spec parser now falls back to landscape when missing an underscore (e.g., a4l to a4_l).

• Pipeline execution now fails with an error if stage arguments cannot be parsed, rather than warning and proceeding.

• Improved error messages for corrupted or invalid PDF files.

• clip operation: enclose page content in a clipping rectangle

• Absolute rectangle specifications for crop and clip

• montage operation: impose multiple pages onto a grid layout, useful for contact sheets and N-up handouts. Supports configurable grid size, canvas, margins, gutters, and aspect ratio control (fit=contain|fill).

• booklet operation: reorder and impose pages for duplex booklet printing. Automatically pads to a multiple of 4, supports signature-based chunking (sig=N), custom canvas size, and right-to-left binding (rtl=true).

• --version now displays core dependencies (pikepdf, libqpdf, python) and optional dependencies, along with a docs URL

• crop with an invalid spec now prints a clean error message instead of a traceback

• --version no longer crashes if pikepdf is not installed

• Bug fix: render should no longer save a PDF to the pattern file

• Inline pipeline substitution using new JOB and DONE keywords, documented under pdftl help pipeline

• Bump pikepdf required version to 10.3.0 to enable better pdftk compatibility

• Compatibility:

  • Default to encrypt_aes128 encryption, the strongest which is pdftk compatible
  • pdftl now passes the vendored pdftk-java test suite

• Shell completion improvements and optimization

• Respect XDG environment variables for cache and plugin directory on non-Windows

• Performance improvements, particularly in cat and dump_data

• API now accepts io.BytesIO as PDF inputs (as well as pathlib.Path, etc)

• Resolved issue with outlines (contents) when using cat with named destinations

• Shell completion for bash, zsh and powershell

• Performance improvements for cat, shuffle, and rotate operations, especially regarding hyperlink handling.

• Invalid page arguments (e.g., requesting page 10 of a 5-page PDF) now return descriptive error messages instead of crashing.

• delete: Now properly removes page resources, resulting in smaller output files.

• generate_fdf: Fixed an issue where generation could fail or produce invalid output when handling None values or specific binary string formats.

• The option parser (used in add_text, etc.) now correctly handles unbalanced quotes by raising a descriptive error, preventing data corruption.

• Replaced regex parsing with a state machine to prevent application hangs (ReDoS) when processing malformed quoted strings.

• Now using defusedxml to avoid xml.etree.ElementTree vulnerabilities.

• Ensure passwords are redacted from all logging calls.

• README.md, NOTICE.md and CHANGELOG.md updated/corrected

• mutate_content operation: mutate page content streams using a user-supplied Python script

• update_info: Support setting PdfID0 to "RESET" to force valid ID regeneration on save.

• replacement_font option: change the font used for user text in forms

• --version includes optional dependency versions

• Documentation updated for dump_data, dump_annots, and dump_data_fields to better explain output formats and pdftk compatibility.

• Compatibility: generate_fdf fixed and replacement_font implemented. Now passes the pdftk-java tests using bleeding edge pikepdf.

• startup performance improvements: lazy loading, and working around heavy Rich formatting for default help text

• Pipeline: Read-only operations (like dump_text) now implicitly pass the input PDF to the next stage, fixing chained commands (e.g., dump_text --- cat).

• Handling of features with missing dependencies should be more consistent.

• Official support for Python 3.14

• More comprehensive vendor_tests/pdftk-java test suite, from pdftk-java. This test suite only is licensed under the GPL 2, see NOTICE.md.

• no_encrypt_metadata option: do not encrypt metadata. Only supported by AES encryption methods.

• COMPATIBILITY.md: compatibility notes

• dump_data_fields now extracts tooltips (FieldNameAlt) and default values (FieldValueDefault).

• API: inputs now support pathlib.Path objects directly.

• Encryption method now defaults to encrypt_aes128 if user_pw or owner_pw are passed, instead of not encrypting. This is more similar to pdftk.

• attach_files is now an operation, not an output option (pdftk compatibility)

• exit codes better aligned with pdftk's exit codes

• CLI: stricter argument parsing now raises DuplicateArgumentError if keywords are repeated.

• cat should now properly handle forms.

• fill_form compatibility fixes

• dump_data_fields_utf8: fixed crash/output issues.

• set_info: added validation for page label indices and better error handling for rotation/mediabox.

• automated pdftk compatibility testing using third party php test suite.

• drop_xfa output option to drop XFA form data (pdftk compatibility)

• render operation: rasterize pages

• move, update_info, update_info_utf8 now accept instructions from a JSON "at-file" using @filename.json in place of CLI arguments

• dump_data gives JSON output via the json keyword

• extensibility: add custom operations by putting Python files in ~/.config/pdftl/operations (*nix) or %APPDATA%\pdftl\config (windows)

• bug in add_pages.py when a page has an integer key-value

• more comprehensive handling of the five PDF page boxes for dump_data and update_info

• drop_info and drop_xmp output options should now work as claimed

• flatten reimplemented for robustness

• move operation: move pages within a PDF file

• place operation: shift, scale and/or spin content within the page

• Now handles compound page specifications like 1,3-4,7-end

• Improved API documentation generation

• spin operation. See place for this functionality.

• add_text features:

  • source metadata variables (source_filename, source_page, etc)

  • Bates stamping variable features, e.g. DEF-{page+120:06d} produces DEF-000121, DEF-000122, ...

• insert operation: insert blank pages

• Broken link in README.md

• API with fluent and functional interfaces
• docs/api_tutorial.md and auto-generated API docs

• Renamed operation: list_files is now dump_files

• Fixed bug preventing parsing of the page specification "right"

• README.md corrected, and "platform" badge added

• Get help by tag with pdftl help tag:<tagname>

• dump_signatures: view and validate PDF signatures

• PDF signature output options:

  • sign_cert <file> Path to certificate PEM

  • sign_field <name> Signature field name (default: Signature1)

  • sign_key <file> Path to private key PEM

  • sign_pass_env <var> Environment variable with sign_cert passphrase

  • sign_pass_prompt Prompt for sign_cert passphrase

• dump_layers: dump PDF optional content groups (OCGs), a.k.a. "layers"

• performance improvements for cat

• crop: added fit and fit-group

• artwork

• extended NOTICE.md: acknowledge pikepdf/qpdf and pypdfium2

• Windows testing

• performance improvements (lazy-loaded imports)

• help tweaks: add sources for non-operations; more help topic aliases

• readthedocs integration and docs generation

• Improved help text

• codecov integration
• PyPI publish integration

• Initial public release of pdftl.
• Operations:
  • add_text Add user-specified text strings to PDF pages
  • background Use a 1-page PDF as the background for each page
  • burst Split a single PDF into individual page files
  • cat Concatenate pages from input PDFs into a new PDF
  • chop Chop pages into multiple smaller pieces
  • crop Crop pages
  • delete Delete pages from an input PDF
  • delete_annots Delete annotation info
  • dump_annots Dump annotation info
  • dump_data Metadata, page and bookmark info (XML-escaped)
  • dump_data_annots Dump annotation info in pdftk style
  • dump_data_fields Print PDF form field data with XML-style escaping
  • dump_data_fields_utf8 Print PDF form field data in UTF-8
  • dump_data_utf8 Metadata, page and bookmark info (in UTF-8)
  • dump_dests Print PDF named destinations data to the console
  • dump_text Print PDF text data to the console or a file
  • fill_form Fill a PDF form
  • filter Do nothing. (The default if <operation> omitted.)
  • generate_fdf Generate an FDF file containing PDF form data
  • inject Inject code at start or end of page content streams
  • list_files List file attachments
  • modify_annots Modify properties of existing annotations
  • multibackground Use multiple pages as backgrounds
  • multistamp Stamp multiple pages onto an input PDF
  • normalize Reformat page content streams
  • optimize_images Optimize images
  • replace Regex replacement on page content streams
  • rotate Rotate pages in a PDF
  • shuffle Interleave pages from multiple input PDFs
  • spin Spin page content in a PDF
  • stamp Stamp a 1-page PDF onto each page of an input PDF
  • unpack_files Unpack file attachments
  • update_info Update PDF metadata
  • update_info_utf8 Update PDF metadata from dump_data_utf8 instructions
• Output options:
  • allow <perm>... Specify permissions for encrypted files
  • attach_files <file>... Attach files to the output PDF
  • compress (default) Compress output file streams
  • drop_info Discard document-level info metadata
  • drop_xmp Discard document-level XMP metadata
  • encrypt_128bit Use 128 bit encryption (obsolete, maybe insecure)
  • encrypt_40bit Use 40 bit encryption (obsolete, highly insecure)
  • encrypt_aes128 Use 128 bit AES encryption (maybe obsolete)
  • encrypt_aes256 Use 256 bit AES encryption
  • flatten Flatten all annotations
  • keep_final_id Copy final input PDF's ID metadata to output
  • keep_first_id Copy first input PDF's ID metadata to output
  • linearize Linearize output file(s)
  • need_appearances Set a form rendering flag in the output PDF
  • output <file> The output file path, or a template for 'burst'
  • owner_pw <pw> Set owner password and encrypt output
  • uncompress Disables compression of output file streams
  • user_pw <pw> Set user password and encrypt output
  • verbose Turn on verbose output