pdfsink 0.1.2 → 0.2.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 0406a9057adb9c3470182b98d6600e8bf239227e792e4dd1f9560542ab5056bb
-  data.tar.gz: e048fcdd3eaf050bcd91cd289930822ff17a28c6603b45e22c5a857982374a6f
+  metadata.gz: 6097674e65f4c20f1f715625ff7e45ba1de03af6c5a42e019fc52b4c7b471897
+  data.tar.gz: 6eaa5deed917aa12a71f1dce9918b814da3c2078eb2bcea530a82e8d62e21400
 SHA512:
-  metadata.gz: '098ac8d04c6585bd7355beb2a5aec26d6d4e625e49248baa00907161ce1ddb870be0012ede8e67bf327afb2dcf3a03dacab1ee32076dea1ecb737176d1c32f28'
-  data.tar.gz: 9679286de91e56311407459e52f3214171b7977ec06ba377c943f5b901b1f1973bc7ac7bd7d95cd3ae34c6694cad227f326be5f76f592de8346294ca54bd5571
+  metadata.gz: fbaf27f8a1e2fd4edb52c244b6e8a98f66eb35e71d1652bbaea5ff420f29fd32b7280defdb6290e40a9fec418882a8600ef152bcde7995055ebc071463f05c4e
+  data.tar.gz: 36566d61641837c8a85bfd23f468f184ab3bcd6e7e9e5e1a6ee5f0ebc30453b607423d8c4baf283b5057e024cc39c1753e101284c796f85ce367357c24f8652d

data/README.md

@@ -106,10 +106,30 @@ page.search(/total:\s*\$\d+/i)
 # => [{"text" => "Total: $420", "x0" => ..., ...}]
 ```
 
+### Read every page in one run
+
+Each page method runs the pdfsink-rs binary for that page only. To read
+every page, `load_all` runs it once for the whole document:
+
+```ruby
+doc = Pdfsink.open("report.pdf")
+doc.load_all                     # one run; returns doc.pages
+doc.pages.each do |page|
+  page.extract_text              # no further runs
+  page.tables                    # no further runs (same strategy as load_all)
+end
+
+doc.load_all(strategy: :text)    # load tables found with another strategy
+```
+
+Use the page methods when you need a few pages, and `load_all` when you
+need all of them. `load_all` loads text and tables only; `extract_words`,
+`objects`, `links` and `search` still run the binary for their page.
+
 ### Binary version
 
 ```ruby
-Pdfsink.version   # => "0.2.9"
+Pdfsink.version   # => "0.2.20+aim.1"
 ```
 
 ## Rails Integration

data/lib/pdfsink/cli.rb

@@ -30,7 +30,7 @@ module Pdfsink
       # Override the resolved binary path (mainly for tests).
       attr_writer :binary
 
-      # The pdfsink-rs version string, e.g. "pdfsink-rs 0.2.9".
+      # The pdfsink-rs version string, e.g. "0.2.20+aim.1".
       #
       # The CLI has no version subcommand, so this reports the crate version
       # the gem was built against.
@@ -51,11 +51,14 @@ module Pdfsink
 
       # Extracted text for a single page.
       #
+      # The binary always writes UTF-8. Open3 tags its output with the process
+      # locale, which is wrong under a C locale, so the tag is set here.
+      #
       # @param path [String]
       # @param page [Integer] 1-based page number
-      # @return [String]
+      # @return [String] UTF-8
       def text(path, page)
-        run("text", path, page.to_s)
+        run("text", path, page.to_s).force_encoding(Encoding::UTF_8)
       end
 
       # Words with positions for a single page.
@@ -94,6 +97,19 @@ module Pdfsink
         run_json("table", path, page.to_s, strategy)
       end
 
+      # Document metadata plus the text and largest table of every page, in
+      # one run. Same payload as {info}, with "text" and "table" added to
+      # each entry in "pages". "text" has no trailing newline, unlike {text}.
+      #
+      # @param path [String]
+      # @param strategy [String] one of "lines", "lines_strict", "text", "explicit"
+      # @return [Hash]
+      # @raise [ParseError] if the binary prints nothing (too old to have +pages+)
+      def pages(path, strategy)
+        run_json("pages", path, strategy) or
+          raise ParseError, "pdfsink-rs pages printed nothing; the binary may be older than #{PDFSINK_RS_VERSION}"
+      end
+
       private
 
       def find_binary

data/lib/pdfsink/document.rb

@@ -5,7 +5,8 @@ module Pdfsink
   #
   # Opening is cheap: the path is validated and the document's +info+ payload
   # (page count and per-page metadata) is fetched lazily on first access.
-  # {Page} objects are created on demand and memoized.
+  # {Page} objects are created on demand and memoized. {#load_all} reads
+  # every page's text and table in one run.
   #
   # @example
   #   doc = Pdfsink::Document.open("report.pdf")
@@ -77,6 +78,30 @@ module Pdfsink
       pages.each(&block)
     end
 
+    # Read the text and largest table of every page in one run of the
+    # binary. Afterwards {Page#extract_text}, and {Page#tables} with the same
+    # strategy, return without running it again. Use this when you need
+    # every page; for a few pages, the per-page methods are faster.
+    #
+    # Each call parses the whole document again, so a second strategy needs a
+    # second call. {Page#extract_words}, {Page#objects}, {Page#links} and
+    # {Page#search} are not preloaded.
+    #
+    # @param strategy [Symbol, String, nil] table-detection strategy
+    # @return [Array<Page>]
+    def load_all(strategy: nil)
+      strategy = TableStrategy.resolve(strategy)
+      payload  = Cli.pages(path, strategy)
+      entries  = payload["pages"]
+
+      @info = payload.merge("pages" => entries.map { |entry| entry.except("text", "table") })
+      entries.each.with_index(1) do |entry, number|
+        # The text command prints a trailing newline; pages does not.
+        page(number).preload(text: "#{entry['text']}\n", strategy: strategy, table: entry["table"])
+      end
+      pages
+    end
+
     def inspect
       "#<Pdfsink::Document path=#{path.inspect} pages=#{page_count}>"
     end

data/lib/pdfsink/page.rb

@@ -3,10 +3,11 @@
 module Pdfsink
   # A single page of a {Document}.
   #
-  # Each accessor shells out to the pdfsink-rs binary for that page; results
-  # are cached so repeated reads don't re-spawn the process. Page-level
-  # metadata (dimensions, rotation, bbox, object counts) comes from the
-  # document's +info+ payload and needs no extra spawn.
+  # Each accessor runs the pdfsink-rs binary for this page only; results are
+  # cached so repeated reads don't run it again. {Document#load_all} fills
+  # the text and tables of every page in one run. Page-level metadata
+  # (dimensions, rotation, bbox, object counts) comes from the document's
+  # +info+ payload and needs no extra run.
   #
   # @example
   #   doc  = Pdfsink::Document.open("report.pdf")
@@ -25,6 +26,7 @@ module Pdfsink
       @document = document
       @number   = number
       @meta     = meta
+      @tables   = {}
     end
 
     # @return [Float] page width in PDF points
@@ -83,7 +85,22 @@ module Pdfsink
     # @param strategy [Symbol, String, nil] table-detection strategy
     # @return [Array<Array>, nil] rows of cells
     def tables(strategy: nil)
-      Cli.table(path, number, TableStrategy.resolve(strategy))
+      strategy = TableStrategy.resolve(strategy)
+      return @tables[strategy] if @tables.key?(strategy)
+
+      @tables[strategy] = Cli.table(path, number, strategy)
+    end
+
+    # Store text and a table read by {Document#load_all}.
+    #
+    # @api private
+    # @param text [String]
+    # @param strategy [String] a resolved table strategy
+    # @param table [Array<Array>, nil]
+    # @return [void]
+    def preload(text:, strategy:, table:)
+      @extract_text = text
+      @tables[strategy] = table
     end
 
     def inspect

data/lib/pdfsink/version.rb

@@ -1,10 +1,10 @@
 # frozen_string_literal: true
 
 module Pdfsink
-  VERSION = "0.1.2"
+  VERSION = "0.2.0"
 
   # Version of the pdfsink-rs crate this gem builds and wraps. Built from the
   # matching git tag (v#{PDFSINK_RS_VERSION}), not crates.io -- the crate carries
   # a [patch.crates-io] fix that cargo strips on publish.
-  PDFSINK_RS_VERSION = "0.2.12"
+  PDFSINK_RS_VERSION = "0.2.20+aim.1"
 end

data/lib/pdfsink.rb

@@ -73,7 +73,7 @@ module Pdfsink
 
     # The version of the underlying pdfsink-rs binary the gem was built with.
     #
-    # @return [String] e.g. "0.2.9"
+    # @return [String] e.g. "0.2.20+aim.1"
     def version
       Cli.version
     end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: pdfsink
 version: !ruby/object:Gem::Version
-  version: 0.1.2
+  version: 0.2.0
 platform: ruby
 authors:
 - Accountaim
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2026-07-17 00:00:00.000000000 Z
+date: 2026-10-24 00:00:00.000000000 Z
 dependencies: []
 description: |
   A Ruby gem that wraps the pdfsink-rs CLI, a fast pure-Rust PDF extraction