imprimatur 0.7.0-x86_64-linux-gnu → 0.8.1-x86_64-linux-gnu

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 61d8e721195a99877afa30f44305b3e5a188b2f703a7ec91ca13b97d7a81f554
-  data.tar.gz: e6ed13beffae7614cc2d869c2a6cf0a25b3fdc8f764d6b8a48408f85d06059be
+  metadata.gz: f633e6ff9c73538d14383925c761a1a06d19a5252b70b3b0351d1875696755b8
+  data.tar.gz: f098e38ba9eb17115703bcaa36a1621872392325721b6a90e028f4adaf39a368
 SHA512:
-  metadata.gz: cf47de8bfd6e94e089b11e44af6c920de1100f4ebba9be22b98dfb27ecf844d215698f0baa12e1e709e2d4c8bc4ce1d78f992218a94096509396408974f4c679
-  data.tar.gz: 972bdda1bc9679629acb56a82a7d9baf15231c4f06df5bde341dded3f272ad8fe8ba3ffa7d84f2316424a4ae8d1ad9c271b904481562ff2c990d3b499e8ab1ed
+  metadata.gz: bc99ae942f2a2c1ed48b1ebcdacd9cede4227137eb87407aa6d77da9332d96a7376c419a593a857d6d237b4961ec72c167f68c69920fd3e4f21b929d382c647b
+  data.tar.gz: 30c68466652b1260f8819df9eda7a924e6d7246514f871e3ef09077000e692c70ac7cbbcd87a335bcc1f3b462ef1e2e18528cfd95c72b74b62bbd693b75a22e8

data/CHANGELOG.md

@@ -11,6 +11,26 @@
 - Allow a bounded exit grace when macOS cannot inspect an exiting worker;
   persistent monitor failures still terminate and reap a live worker.
 
+## [0.8.1](https://github.com/seuros/imprimatur/compare/v0.8.0...v0.8.1) (2026-10-29)
+
+
+### Bug Fixes
+
+* **pdf:** recover timeout neighbours within the overall deadline ([#16](https://github.com/seuros/imprimatur/issues/16)) ([e86fbbf](https://github.com/seuros/imprimatur/commit/e86fbbfef85f2c645090c43359c0744962e951a1))
+* **pdf:** step timeout recovery so a Ruby wakeup cannot restart it ([75c3894](https://github.com/seuros/imprimatur/commit/75c3894c8a791c5ef4af562dc99de4e745770ad3))
+
+## [0.8.0](https://github.com/seuros/imprimatur/compare/v0.7.0...v0.8.0) (2026-10-27)
+
+
+### Features
+
+* **reader:** accept drop_repeated in the Ruby API ([4e537d6](https://github.com/seuros/imprimatur/commit/4e537d641bafd8d3cdb8b1c20d16f4cb766ad073))
+
+
+### Bug Fixes
+
+* **pdf:** read printed page numbers from headers too, not just footers ([c036434](https://github.com/seuros/imprimatur/commit/c0364341ee8fc36be2c97d25fd06da0aa76a4ad6))
+
 ## [0.7.0](https://github.com/seuros/imprimatur/compare/v0.6.1...v0.7.0) (2026-10-26)
 
 

data/README.md

@@ -32,7 +32,7 @@ end
 gem "imprimatur"
 ```
 
-Precompiled gems ship for Linux (x86_64 and aarch64, glibc 2.34+ and musl),
+Precompiled gems ship for Linux (x86_64 and aarch64, glibc 2.34+),
 macOS 13+ (arm64, x86_64) and FreeBSD 15.1 (amd64): nothing to compile.
 Anywhere else, the source gem compiles the extension on install, with the
 toolchain from [rustup.rs](https://rustup.rs) (1.99+).
@@ -384,6 +384,28 @@ after metadata was read preserves completed pages and explicitly marks the
 unreached pages; an expired overall deadline marks them `skipped`. Cancellation
 still raises rather than returning a partial successful call.
 
+For timeout-neighbour recovery, opt in to a shorter worker-attempt budget:
+
+```ruby
+Imprimatur.pdf_pages(bytes, timeout: 300, batch_timeout: 60)
+```
+
+Here `timeout:` is the **overall** deadline, including recovery; `batch_timeout:`
+caps the initial worker and each singleton retry. Completed pages survive a
+batch timeout. Unread later neighbours are tried once before the first missing
+page (which may have stalled the worker). Recovery is best effort, never an
+extension of the overall deadline. Output-limit skips are not retried, and all
+attempts share `max_output_bytes:`. Input, depth, process-memory and per-page
+limits still apply; `max_inflated:` remains per worker/recovery attempt, not a
+cumulative decompression allowance.
+
+With `batch_timeout:` enabled, an explicit finite `timeout:` replaces the
+ordinary 60-second process ceiling for this call; `timeout: nil` retains that
+60-second ceiling. Both budgets must be positive; zero overall time still fails
+immediately. Omitting `batch_timeout:` preserves the existing behaviour. Rust
+callers can use `pages_isolated_recovering` with an explicit attempt duration
+and overall `Limits`/`ProcessLimits`.
+
 This is resource containment, **not a filesystem/network security sandbox**.
 The raw crate `info`, `pages` and `render_page` APIs remain in-process;
 `render_page` cannot interrupt Hayro.

data/lib/imprimatur/pdf_reader.rb

@@ -11,8 +11,9 @@ module Imprimatur
 
     def self.pages(document, first_page: 1, last_page: 0xFFFFFFFF, max_pages: 500,
                    max_page_bytes: 256 * 1024, max_output_bytes: 8 * 1024 * 1024,
-                   memory_bytes: 1024 * 1024 * 1024, max_bytes: nil, **limits)
-      options = [first_page, last_page, max_pages, max_page_bytes, max_output_bytes]
+                   memory_bytes: 1024 * 1024 * 1024, max_bytes: nil, batch_timeout: nil, **limits)
+      options = [first_page, last_page, max_pages, max_page_bytes, max_output_bytes,
+                 batch_timeout && Float(batch_timeout)]
       JSON.parse(Native.pdf_pages(Reader.read(document, max_bytes), Reader.limits(max_bytes:, **limits),
                                   options, RENDER_WORKER, memory_bytes),
                  symbolize_names: true)

data/lib/imprimatur/reader.rb

@@ -41,8 +41,8 @@ module Imprimatur
     end
 
     # @api private
-    def self.limits(timeout: 60, max_bytes: nil, max_inflated: nil)
-      [max_bytes, max_inflated, timeout && Float(timeout)]
+    def self.limits(timeout: 60, max_bytes: nil, max_inflated: nil, drop_repeated: false)
+      [max_bytes, max_inflated, timeout && Float(timeout), drop_repeated ? true : false]
     end
 
     # A path or an IO is read no further than +max_bytes+ (one byte past it

data/lib/imprimatur/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Imprimatur
-  VERSION = '0.7.0'
+  VERSION = '0.8.1'
 end

data/lib/imprimatur.rb

@@ -72,6 +72,8 @@ module Imprimatur
     # @param timeout [Numeric, nil] seconds the read may take (60), +nil+ for none
     # @param max_bytes [Integer, nil] the largest document accepted (512 MiB)
     # @param max_inflated [Integer, nil] decompressed bytes allowed in all (1 GiB)
+    # @param drop_repeated [Boolean] PDF only: drop text repeated at the same
+    #   spot on most pages, anywhere on the page (watermarks); off by default
     # @return [String] Markdown, UTF-8
     # @raise [NoTextLayer] a PDF of scanned pages: run OCR instead
     # @raise [PasswordRequired] a PDF that needs a password to open
@@ -113,6 +115,8 @@ module Imprimatur
     # image count (nil when unknown), truncation flag and failure reason.
     # Takes +first_page:+, +last_page:+, +max_pages:+ (500),
     # +max_page_bytes:+ (256 KiB), +max_output_bytes:+ (8 MiB), and read limits.
+    # Optional +batch_timeout:+ caps each worker attempt; +timeout:+ then bounds
+    # the whole call, including one-pass recovery of timeout-unread neighbours.
     # @return [Hash] +page_count:+, +encrypted:+ and +pages:+
     def pdf_pages(document, ...) = PdfReader.pages(document, ...)
 

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: imprimatur
 version: !ruby/object:Gem::Version
-  version: 0.7.0
+  version: 0.8.1
 platform: x86_64-linux-gnu
 authors:
 - Abdelkader Boudih