Add option to set page.date automatically with the file creation date

README.md

@@ -25,6 +25,33 @@ last-modified-at:
     date-format: '%d-%b-%y'
 ```
 
+You can also set `override-date: true` to have "last-modified-at" determine the datetime the file was first commited and use that to automatically set the post's `date` field.  Will use `ctime` if there's no git information.
+
+For sites with lots of documents using `last_modified_at`, there may be render
+performance improvement via:
+
+```yml
+plugins:
+  - jekyll-last-modified-at
+
+last-modified-at:
+    use-git-cache: true
+```
+
+If `use-git-cache` is `false` (the default), every committed file using
+`last_modified_at` will generate a separate spawned process to check the git log
+for time data. So if you have 10 documents, this will result in 10 spawned calls.
+
+If `use-git-cache` is `true`, a single spawned process is generated that reads
+the entire git log history and caches the time data. This cache is then read
+from during the rest of the site generation process. So if you have 10 (or 1000)
+documents, this will result in 1 spawned call. The cache is flushed on site
+reset, allowing for a long-lived server to correctly reflect `last_modified_at`
+of files modified and committed while it has been running.
+
+Note: there may be performance issues for repositories with very large
+histories, in which case the default behavior is likely preferred.
+
 ## Usage
 
 There are a few ways to use this gem.

lib/jekyll-last-modified-at.rb

@@ -9,8 +9,5 @@ module LastModifiedAt
     autoload :Executor, 'jekyll-last-modified-at/executor'
     autoload :Determinator, 'jekyll-last-modified-at/determinator'
     autoload :Git, 'jekyll-last-modified-at/git'
-
-    PATH_CACHE = {} # rubocop:disable Style/MutableConstant
-    REPO_CACHE = {} # rubocop:disable Style/MutableConstant
   end
 end

lib/jekyll-last-modified-at/determinator.rb

@@ -3,49 +3,71 @@
 module Jekyll
   module LastModifiedAt
     class Determinator
-      attr_reader :site_source, :page_path
+      @repo_cache = {}
+      @last_mod_cache = {}
+      @first_mod_cache = {}
+      class << self
+        # attr_accessor so we can flush externally
+        attr_accessor :repo_cache
+        attr_accessor :last_mod_cache
+        attr_accessor :first_mod_cache
+      end
+
+      attr_reader :site_source, :page_path, :use_git_cache
       attr_accessor :format
 
-      def initialize(site_source, page_path, format = nil)
-        @site_source = site_source
-        @page_path   = page_path
-        @format      = format || '%d-%b-%y'
+      def initialize(site_source, page_path, format = nil, use_git_cache = false, first_time = false) # rubocop:disable Style/OptionalBooleanParameter
+        @site_source   = site_source
+        @page_path     = page_path
+        @format        = format || '%d-%b-%y'
+        @use_git_cache = use_git_cache
+        @first_time    = first_time
       end
 
       def git
-        return REPO_CACHE[site_source] unless REPO_CACHE[site_source].nil?
+        return self.class.repo_cache[site_source] unless self.class.repo_cache[site_source].nil?
 
-        REPO_CACHE[site_source] = Git.new(site_source)
-        REPO_CACHE[site_source]
+        self.class.repo_cache[site_source] = Git.new(site_source)
+        self.class.repo_cache[site_source]
       end
 
       def formatted_last_modified_date
-        return PATH_CACHE[page_path] unless PATH_CACHE[page_path].nil?
+        last_modified_at_time.strftime(@format)
+      end
+
+      def formatted_first_modified_date
+        first_modified_at_time.strftime(@format)
+      end
+
+      def first_modified_at_time
+        return self.class.first_mod_cache[page_path] unless self.class.first_mod_cache[page_path].nil?
 
-        last_modified = last_modified_at_time.strftime(@format)
-        PATH_CACHE[page_path] = last_modified
-        last_modified
+        raise Errno::ENOENT, "#{absolute_path_to_article} does not exist!" unless File.exist? absolute_path_to_article
+
+        self.class.first_mod_cache[page_path] = Time.at(first_modified_at_unix.to_i)
+        self.class.first_mod_cache[page_path]
+      end
+
+      def first_modified_at_unix
+        if git.git_repo?
+          first_commit_date = git.first_commit_date(relative_path_from_git_dir, use_git_cache)
+          first_commit_date.nil? || first_commit_date.empty? ? ctime(absolute_path_to_article) : first_commit_date
+        else
+          ctime(absolute_path_to_article)
+        end
       end
 
       def last_modified_at_time
+        return self.class.last_mod_cache[page_path] unless self.class.last_mod_cache[page_path].nil?
+
         raise Errno::ENOENT, "#{absolute_path_to_article} does not exist!" unless File.exist? absolute_path_to_article
 
-        Time.at(last_modified_at_unix.to_i)
+        self.class.last_mod_cache[page_path] = Time.at(last_modified_at_unix.to_i)
       end
 
       def last_modified_at_unix
         if git.git_repo?
-          last_commit_date = Executor.sh(
-            'git',
-            '--git-dir',
-            git.top_level_directory,
-            'log',
-            '-n',
-            '1',
-            '--format="%ct"',
-            '--',
-            relative_path_from_git_dir
-          )[/\d+/]
+          last_commit_date = git.last_commit_date(relative_path_from_git_dir, use_git_cache)
           # last_commit_date can be nil iff the file was not committed.
           last_commit_date.nil? || last_commit_date.empty? ? mtime(absolute_path_to_article) : last_commit_date
         else
@@ -54,11 +76,27 @@ def last_modified_at_unix
       end
 
       def to_s
-        @to_s ||= formatted_last_modified_date
+        if @first_time
+          @to_s ||= formatted_first_modified_date
+        else
+          @to_s ||= formatted_last_modified_date
+        end
       end
 
       def to_liquid
-        @to_liquid ||= last_modified_at_time
+        if @first_time
+          @to_liquid ||= first_modified_at_time
+        else
+          @to_liquid ||= last_modified_at_time
+        end
+      end
+
+      def to_time
+        to_liquid
+      end
+
+      def strftime(*args)
+        return to_liquid().strftime(*args)
       end
 
       private
@@ -79,6 +117,10 @@ def relative_path_from_git_dir
       def mtime(file)
         File.mtime(file).to_i.to_s
       end
+
+      def ctime(file)
+        File.ctime(file).to_i.to_s
+      end
     end
   end
 end

lib/jekyll-last-modified-at/git.rb

@@ -8,6 +8,8 @@ class Git
       def initialize(site_source)
         @site_source = site_source
         @is_git_repo = nil
+        @lcd_cache = {}
+        @lce_cache = {}
       end
 
       def top_level_directory
@@ -33,6 +35,80 @@ def git_repo?
                          false
         end
       end
+
+      def last_commit_date(path, use_git_cache = false) # rubocop:disable Style/OptionalBooleanParameter
+        if use_git_cache
+          build_cache if @lcd_cache.empty?
+          @lcd_cache[path]
+        else
+          Executor.sh(
+            'git',
+            '--git-dir',
+            top_level_directory,
+            'log',
+            '-n',
+            '1',
+            '--format="%ct"',
+            '--',
+            path
+          )[/\d+/]
+        end
+      end
+
+      def first_commit_date(path, use_git_cache = false) # rubocop:disable Style/OptionalBooleanParameter
+        if use_git_cache
+          build_cache if @lce_cache.empty?
+          @lce_cache[path]
+        else
+          Executor.sh(
+            'git',
+            '--git-dir',
+            top_level_directory,
+            'log',
+            '--format="%ct"',
+            '--',
+            path
+          ).split("\n")[-1][/\d+/]
+        end
+      end
+
+      private
+
+      # generates hash of `path => unix time stamp (string)`
+      def build_cache
+        # example output:
+        #
+        # %these-files-modified-at:1621042992
+        #
+        # Dockerfile.production
+        # %these-files-modified-at:1621041929
+        #
+        # assets/css/style.52513a5600efd4015668ccb9b702256e.css
+        # assets/css/style.52513a5600efd4015668ccb9b702256e.css.gz
+        lines = Executor.sh(
+          'git',
+          '--git-dir',
+          top_level_directory,
+          'log',
+          '--name-only',
+          '--date=unix',
+          '--pretty=%%these-files-modified-at:%ct'
+        )
+
+        timestamp = nil
+        lines.split("\n").each do |line|
+          next if line.empty?
+
+          if line.start_with?('%these-files-modified-at:')
+            # new record
+            timestamp = line.split(':')[1]
+            next
+          end
+
+          @lcd_cache[line] = timestamp unless @lcd_cache.key?(line)
+          @lce_cache[line] = timestamp
+        end
+      end
     end
   end
 end

lib/jekyll-last-modified-at/hook.rb

@@ -6,8 +6,15 @@ module Hook
       def self.add_determinator_proc
         proc { |item|
           format = item.site.config.dig('last-modified-at', 'date-format')
-          item.data['last_modified_at'] = Determinator.new(item.site.source, item.path,
-                                                           format)
+          use_git_cache = item.site.config.dig('last-modified-at', 'use-git-cache')
+          item.data['last_modified_at'] = Determinator.new(item.site.source, item.relative_path,
+                                                           format, use_git_cache)
+          if item.site.config.dig('last-modified-at', 'override-date')
+            # The "date" field will be converted to a string first by Jekyll and it must be
+            # in the format given below: https://jekyllrb.com/docs/variables/#page-variables
+            item.data['date'] = Determinator.new(item.site.source, item.relative_path,
+                                                 '%Y-%m-%d %H:%M:%S %z', use_git_cache, true)
+          end
         }
       end