Re: [Rails] What's the best way to approach reading and parse large XLSX files?

[Walter Lee Davis]

On Oct 10, 2013, at 4:36 PM, Monserrat Foster wrote:

> Hello, I'm developing an app that basically, receives a 10MB or less XLSX files with +30000 rows or so, and another XLSX file with about 200rows, I have to read one row of the smallest file, look it up on the largest file and write data from both files to a new one.

Wow. Do you have to do all this in a single request?

You may want to look at Nokogiri and its SAX parser. SAX parsers don't care about the size of the document they operate on, because they work one node at a time, and don't load the whole thing into memory at once. There are some limitations on what kind of work a SAX parser can perform, because it isn't able to see the entire document and "know" where it is within the document at any point. But for certain kinds of problems, it can be the only way to go. Sounds like you may need something like this.

Walter

>
> I just did a test reading a few rows from the largest file using ROO (Spreadsheet doesn't support XSLX and Creek look good but I can't find a way to read row by row)
> and it basically made my computer crash, the server crashed, I tried rebooting it and it said It was already started, anyway, it was a disaster.
>
> So, my question was, is there gem that works best with large XLSX files or is there another way to approach this withouth crashing my computer?
>
> This is what I had (It's very possible I'm doing it wrong, help is welcome)
> What i was trying to do here, was to process the files and create the new XLS file after both of the XLSX files were uploaded:
>
>
> require 'roo'
> require 'spreadsheet'
> require 'creek'
> class UploadFiles < ActiveRecord::Base
> after_commit :process_files
> attr_accessible :inventory, :material_list
> has_one :inventory
> has_one :material_list
> has_attached_file :inventory, :url=>"/:current_user/inventory", :path=>":rails_root/tmp/users/uploaded_files/inventory/inventory.:extension"
> has_attached_file :material_list, :url=>"/:current_user/material_list", :path=>":rails_root/tmp/users/uploaded_files/material_list/material_list.:extension"
> validates_attachment_presence :material_list
> accepts_nested_attributes_for :material_list, :allow_destroy => true
> accepts_nested_attributes_for :inventory, :allow_destroy => true
> validates_attachment_content_type :inventory, :content_type => ["application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"], :message => "Only .XSLX files are accepted as Inventory"
> validates_attachment_content_type :material_list, :content_type => ["application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"], :message => "Only .XSLX files are accepted as Material List"
>
>
> def process_files
> inventory = Creek::Book.new(Rails.root.to_s + "/tmp/users/uploaded_files/inventory/inventory.xlsx")
> material_list = Creek::Book.new(Rails.root.to_s + "/tmp/users/uploaded_files/material_list/material_list.xlsx")
> inventory = inventory.sheets[0]
> scl = Spreadsheet::Workbook.new
> sheet1 = scl.create_worksheet
> inventory.rows.each do |row|
> row.inspect
> sheet1.row(1).push(row)
> end
>
> sheet1.name = "Site Configuration List"
> scl.write(Rails.root.to_s + "/tmp/users/generated/siteconfigurationlist.xls")
> end
> end
>
>
> --
> You received this message because you are subscribed to the Google Groups "Ruby on Rails: Talk" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to rubyonrails-ta...@googlegroups.com.
> To post to this group, send email to rubyonra...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/rubyonrails-talk/bc470d4d-19c4-4969-8ba7-4ead7a35d40c%40googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.

[Monserrat Foster]

A coworker suggested I should use just basic OOP for this, to create a class that reads files, and then another to load the files into memory. Could please point me in the right direction for this (where can I read about it)? I have no idea what's he talking about, as I've never done this before. 

I'll look up nokogiri and SAX

[Martin Streicher]

I highly recommend the RubyXL gem. It opens xlsx files and seems very reliable. I use it all the time. 

 

[Walter Lee Davis]

On Oct 10, 2013, at 4:50 PM, Monserrat Foster wrote:

> A coworker suggested I should use just basic OOP for this, to create a class that reads files, and then another to load the files into memory. Could please point me in the right direction for this (where can I read about it)? I have no idea what's he talking about, as I've never done this before.

How many of these files are you planning to parse at any one time? Do you have the memory on your server to deal with this load? I can see this approach working, but getting slow and process-bound very quickly. Lots of edge cases to deal with when parsing big uploaded files.

Walter

> To view this discussion on the web visit https://groups.google.com/d/msgid/rubyonrails-talk/0325dc87-0649-45fc-9d55-0fbcd8bed0a0%40googlegroups.com.

[Monserrat Foster]

One 30000+ row file and another with just over 200. How much memory should I need for this not to take forever parsing? (I'm currently using my computer as server and I can see ruby taking about 1GB in the task manager when processing this (and it takes forever).

The 30000+ row file is about 7MB, which is not that much (I think) 

[Walter Lee Davis]

On Oct 11, 2013, at 11:30 AM, Monserrat Foster wrote:

> One 30000+ row file and another with just over 200. How much memory should I need for this not to take forever parsing? (I'm currently using my computer as server and I can see ruby taking about 1GB in the task manager when processing this (and it takes forever).
>
> The 30000+ row file is about 7MB, which is not that much (I think)

I have a collection of 1200 XML files, ranging in size from 3MB to 12MB each (they're books, in TEI encoding) that I parse with Nokogiri on a 2GB Joyent SmartMachine to convert them to XHTML and then on to Epub. This process takes 17 minutes for the first pass, and 24 minutes for the second pass. It does not crash, but the server is unable to do much of anything else while the loop is running.

My question here was, is this something that is a self-serve web service, or an admin-level (one-privileged-user-once-in-a-while) type thing? In my case, there's one admin who adds maybe two or three books per month to the collection, and the 40-minute do-everything loop was used only for development purposes -- it was my test cycle as I checked all of the titles against a validator to ensure that my adjustments to the transcoding process didn't result in invalid code. I would not advise putting something like this live against the world, as the potential for DOS is extremely great. Anything that can pull the kinds of loads you get when you load a huge file into memory and start fiddling with it should not be public!

Walter

> To view this discussion on the web visit https://groups.google.com/d/msgid/rubyonrails-talk/ba633f69-5527-4dc1-8518-b6104e414e15%40googlegroups.com.

[Jordon Bedwell]

On Fri, Oct 11, 2013 at 10:30 AM, Monserrat Foster
<monse...@gmail.com> wrote:
> One 30000+ row file and another with just over 200. How much memory should I
> need for this not to take forever parsing? (I'm currently using my computer
> as server and I can see ruby taking about 1GB in the task manager when
> processing this (and it takes forever).
>
> The 30000+ row file is about 7MB, which is not that much (I think)

Check for a memory leak.

[Donald Ziesig]

To view this discussion on the web visit https://groups.google.com/d/msgid/rubyonrails-talk/ba633f69-5527-4dc1-8518-b6104e414e15%40googlegroups.com.

For more options, visit https://groups.google.com/groups/opt_out.

I use a rather indirect route that works fine for me with 15,000 lines and about 26 MB.  I export the file from LibreOffice Calc using csv (Comma separated variables).  Then, in the rails controller I use something like:

require 'csv'

class TheControllerController # ;')

# other controller code

  def upload
    data = CSV.parse(params[:entries].tempfile.read) # from Ruby's CSV class
    for line in data do
      logger.debug "line: #{line.inspect}"
       #each line is an array of strings containing the columns of the one row of the csv file
       #I use these data to populate the appropriate db table / rails model at this point
    end
  end

end

make sure that your routes.db points to this:

  match 'the_controller/upload' => 'the_controller#upload'

from your client machine's command line
 
curl -F entries=@yourdata.csv localhost:3000/the_controller/upload

note that 'entries' in the curl command matches the 'entries' in the param[:entries] in the controller.

If you want to do this from a rails gui form, look at http://guides.rubyonrails.org/form_helpers.html#uploading-files

During testing on my 4-core, 8 GB laptop, processing the really big files take several minutes.  When I have the app on heroku, this causes a timeout so I break up the csv file into multiple sections such that each section takes less than 30 seconds to upload.  By leaving a little 'slack' in the size, I have this automated so it occurs in the background while I am doing other work.

Hope these suggestions help.

Don Ziesig

[Monserrat Foster]

Hi, the files automatically download in .XLSX formats, I can't change them and I can't force the users to change it in order to make my job easier. Thanks for the suggestion. 

curl -F entr...@yourdata.csv localhost:3000/the_controller/upload

[Monserrat Foster]

This is an everyday, initially maybe a couple people at the same time uploading and parsing files to generate the new one, but eventually it will extend to other people, so...

I used a logger and It does retrieve and save the files using the comparation. But it takes forever, like 30min or so in generating the file. 

The process starts as soon as the files are uploaded but it seems to be taking most of the time into opening the file, once it's opened it takes maybe 5min at most to generate the new file.

Do you know where can i find an example on how to read an xlsx file with nokogiri? I can't seem to find one

[Monserrat Foster]

I forgot to say after it reads all rows and writes the file, throws

[1m [35m (600.1ms) [0m  begin transaction

  [1m [36m (52.0ms) [0m   [1mcommit transaction [0m

failed to allocate memory

Redirected to http://localhost:3000/upload_files/110

Completed 406 Not Acceptable in 1207471ms (ActiveRecord: 693.1ms)

[Walter Lee Davis]

On Oct 11, 2013, at 4:33 PM, Monserrat Foster wrote:

> This is an everyday, initially maybe a couple people at the same time uploading and parsing files to generate the new one, but eventually it will extend to other people, so...
>
> I used a logger and It does retrieve and save the files using the comparation. But it takes forever, like 30min or so in generating the file.
> The process starts as soon as the files are uploaded but it seems to be taking most of the time into opening the file, once it's opened it takes maybe 5min at most to generate the new file.
>
> Do you know where can i find an example on how to read an xlsx file with nokogiri? I can't seem to find one

XSLX is just an Excel file expressed in XML. It's no different than parsing any other XML file. First, find a good basic example of file parsing with Nokogiri. http://nokogiri.org/tutorials/searching_a_xml_html_document.html Next, open up your file in a text editor, and look for the elements you want to access. You can use either xpath or css syntax to locate your elements, and Nokogiri allows you to access either attributes or content of any element you can locate. If you run into trouble with all the prefixes that Microsoft like to litter their formats with, you can pass remove_namespaces to clean that right up.

> To view this discussion on the web visit https://groups.google.com/d/msgid/rubyonrails-talk/64a04728-0f8d-4795-9e8f-ca6509e05cbc%40googlegroups.com.

[Ted Kaplan]

Creek is good, I'd also recommend dullard, a gem that I wrote.  Its output format may be more convenient for your case.

https://github.com/thirtyseven/dullard

http://rubygems.org/gems/dullard

-Ted