CODEWITHSUNDEEP
ruby-on-railsruby-on-rails-3activerecord
Lesha Pipiev
Lesha Pipiev

Reputation: 3333

Rails ActiveRecord invalid byte sequence in UTF-8 issue

I am using MSSQL 2005.

I have StringIO object that contains my zip file content.

Here is how I obtain zip binary data: 

    stringio = Zip::ZipOutputStream::write_buffer do |zio|
        Eclaim.find_by_sql("SET TEXTSIZE 67108864")
        zio.put_next_entry("application.xml")
        #zio.write @claim_db[:xml]
        biblio = Nokogiri::XML('<?xml version="1.0" encoding="utf-8"?>' + @claim_db[:xml], &:noblanks)
        zio.write biblio.to_xml

        builder = Nokogiri::XML::Builder.new(:encoding => 'utf-8') do |xml|
            xml.documents {
                docs.where("ext not in (#{PROHIBITED_EXTS.collect{|v| "'#{v}'"}.join(', ')})").each{|doc|
                    zio.put_next_entry("#{doc[:materialtitle_id]}.#{doc[:ext]}")
                    zio.write doc[:efile]

                    xml.document(:id => doc[:materialtitle_id]) {
                        xml.title doc[:title]
                        xml.code doc[:code]
                        xml.filename "#{doc[:materialtitle_id]}.#{doc[:ext]}"
                        xml.extname doc[:ext]
                    }

                }
            }
        end

        zio.put_next_entry("docs.xml")
        zio.write builder.to_xml
    end

    stringio

In my controller I try: 

data.rewind
@claim.docs.create(
    :title => 'Some file',
    :ext => 'zip',
    :size => data.length,
    :receive_date => Time.now,
    :efile => data.sysread
)

But Rails complains invalid byte sequence in UTF-8

Help me pls with it.

Upvotes: 2

Views: 3542

Answers (3)

MetaSkills
MetaSkills

Reputation: 1854

You should break your problem down in constituate parts and troubleshoot smaller units. Remove complexity till you get to the source of the issue. For instance, have you tried just doing a simple Document.create with the attributes in say the console to remove the possibility that your controller code may be buggy? Something like Document.create :efile => File.read('sometiny.zip') and just go from there.

Assuming that works or break, you have a much simpler support request and less noise to issue ratio. Right now I suspect your controller code, not the SQL Server Adapter or the connection mode, as I have both tested to the hilt for simple binary data. Assuming the above does not work, you can then moving to examining smaller components.

For instance, what is the data type of the efile column? Do this in the console to find out, Document.columns_hash['efile'] and look at the @sql_type. Is it something suitable like varbinary(max)?

Moving on from there, what connection mode are you using with the SQL Server Adapter, TinyTDS? By default TinyTDS will convert everything to UTF8 as needed and is really smart about things. I have it tested with everything from binary to many different encodings. BTW, if you are using TinyTDS, did you make sure that you compiled FreeTDS with libiconv so it can do all this properly? You can easily check by doing the following in the console tsql -C assuming you have FreeTDS's binaries in your path. This should output a few lines, look for "iconv library: yes". Also make sure you are running 0.91 or better too!

Lastly, a bit of advice, that SET TEXTSIZE is so wrong there. You only want to do that once per connection. See here https://github.com/rails-sqlserver/activerecord-sqlserver-adapter#configure-connection--app-name

Upvotes: 1

shigeya
shigeya

Reputation: 4912

If the stream is configured as UTF-8 stream, you can't write compressed binary (which may contain any value).

I think, setting data as binary stream before write:

data.force_encoding "ASCII-8BIT"

might help.

Upvotes: 6

Marlin Pierce
Marlin Pierce

Reputation: 10079

What is complaining, ruby, ActiveRecord, or SQL Server? My guess is SQL Server. Make sure the data type of the efile field in the database is a binary BLOB.

Upvotes: 1

Related Questions