Fix method contract bug in detect_format

[turadg]

Calls to detect_format() passed a file, but the method expected a filename.

/Users/turadg/Code/Ruby/treat/lib/treat/entities/entity/buildable.rb:
  214      else
  215        fmt = Treat::Workers::Formatters::
  216:       Readers::Autoselect.detect_format(file,def_fmt)
  217        from_raw_file(file, fmt)
  218      end

/Users/turadg/Code/Ruby/treat/lib/treat/workers/formatters/readers/autoselect.rb:
   13    def self.read(document, options = {})
   14      options = DefaultOptions.merge(options)
   15:     fmt = detect_format(document.file, options[:default_to])
   16      Treat::Workers::Formatters::Readers.
   17      const_get(fmt.cc).read(document,options)
   18    end
   19    
   20:   def self.detect_format(filename, default_to = nil)
   21      
   22      default_to ||= DefaultOptions[:default_to]

3 matches across 2 files

This patch changes the parameter to a file as expected.

[louismullie]

Using the Treat.paths.file setting allows you to change the download directory. Any reason why you specifically need to use Tempfile?

[louismullie]

Sorry, it should be Treat.paths.files.

[turadg]

For one, I didn't know about Treat.paths.files. ;)

But I I also need an in-memory copy of the file contents, so I figure it's best to read from the URL directly to memory, then write to disk so Treat can process it. When there is an in-memory copy of the file, why does Treat need it to be stored on disk as well?

[louismullie]

The idea is to make the download large-file-proof by streaming chunks of the download to disk rather than reading the whole file into memory directly. Eventually, I would want to integrate a similar lazy iteration process within the entity classes as well.

For your use case, I would just do:

doc = Treat::Entities::Document.from_raw_file(file.path)