couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Cliff Williams <cliffywi...@aol.com>
Subject Re: Uploading CSV data to Couchapp
Date Mon, 04 Apr 2011 09:34:21 GMT
David,

I hope you are well.

I think that you have covered your options pretty well.

"- upload the data&  save it into a single "uploaded_csv" document in
CouchDB.  Within CouchDB, detect the presence of a new "uploaded_csv"
document, extract and process the content using Javascript and save it into
multiple "data" records, with appropriate indexing, then dispose of the
"uploaded_csv" document or mark it as "processed".  This seems reasonably
straightforward, but I'm not sure how to detect the presence of a new
"uploaded_csv" document"

This is the approach that I would take.

Couchdb has a quite excellent _changes feed which will notify you (or 
can be set up to notify) in real time on any changes made to specific 
databases.

I personally would use Node.js to monitor the changes feed and process 
your csv files (Javascript and very fast) but you could of course use 
anything (erlang python (Ruby's CSV processing libraries are also quite 
good)).

best regards

Cliff
On 04/04/11 08:58, David Mitchell wrote:
> Hello all,
>
> I'm just about to start on my first (wildly ambitious) Couchapp.  I've had
> quite a bit of Erlang experience, but not for the past couple of years so
> I'm a bit rusty.  I've had a tiny bit of experience with CouchDB via various
> Python scripts, but that's all been treating CouchDB as a "black box"
> database so I've currently got little knowledge of what it can do beyond
> being a document datastore.
>
> Initially, I'm trying to understand my options for uploading CSV files,
> parsing out the content and storing them in CouchDB (one CouchDB record per
> line of CSV content).  While it's reasonably straightforward to do this if I
> was using e.g. Python as a batch load tool, I don't want to go outside
> Javascript for this project if I can avoid it.  The CSV files are anywhere
> from 1k-30k records, with 8-10 fields in each that are straightforward
> timestamps and floating point numbers.
>
> For an old-school Web app with distinct database and app server layers,
> there's a straightforward option - upload the data to a file on the web
> server, then process the data out of the file and load it into your
> database.  Sure there's variations on this approach such as saving data as a
> database blob, but I'm looking for the best CouchApp-specific approach if
> one exists.
>
> Options I can see:
> - upload the data&  save it into a single "uploaded_csv" document in
> CouchDB.  Within CouchDB, detect the presence of a new "uploaded_csv"
> document, extract and process the content using Javascript and save it into
> multiple "data" records, with appropriate indexing, then dispose of the
> "uploaded_csv" document or mark it as "processed".  This seems reasonably
> straightforward, but I'm not sure how to detect the presence of a new
> "uploaded_csv" document (is there a cron equivalent in Couch?) and I'd have
> to track the progress of processing each uploaded CSV file to detect when
> they've been processed into "data" records
> - upload the data&  save it into a single "uploaded_csv" document in
> CouchDB.  Have CouchDB running embedded in an Erlang app, and use Erlang to
> read the "uploaded_csv" data, then send a series of e.g. HTTP PUTs to load
> the data into multiple "data" records in CouchDB.  This just seems ugly to
> me, but I'm pretty confident I could get it working pretty easily
> - upload the data and process it directly into "data" records from a web
> page served from CouchApp.  This seems like it could impact on scalability
> due to having long-running connections between client and server, but at
> least a user would know when their data has been uploaded and processed
> successfully with trivial extra work on my part
> - upload the data, convert it to JSON on the client using clientside
> Javascript, then send it as a set of document uploads (i.e. one document per
> CSV record) from the client to the Couch server.  This would let me parse
> out any bogus CSV content without sending it to the server, but I'll have
> users running browsers on mobile devices and I'm not sure I want to put that
> processing load onto the client
>
> Are there any "recommended" approaches for this type of task?  I suspect
> this question and others I'll ask have probably already been considered and
> dealt with by various experts; if there's a "CouchApp cookbook" with
> recommended solutions for these and other common situations, I'd appreciate
> a pointer to it so I could start to answer my own questions.
>
> Thanks in advance
>
> Dave M.
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message