Support for multipart resources.

[georgeslabreche]

Support for multipart resources.

[georgeslabreche]

@roll this is already supported unless I am missing something? Please have a look at the README where I document how to instanciate this datapackage with multipart resources. We can also add and remove Resources.

[roll]

@georgeslabreche This means having an individual resource split into a few parts:

• data-part1.csv -> headers/row1/row2
• data-part2.csv -> row3/row4/etc

http://frictionlessdata.io/specs/data-resource/#data-in-multiple-files

---

This is the lowest priority feature from the specs-v1.

Now it's implemented only for Python - https://github.com/frictionlessdata/datapackage-py/blob/master/datapackage/resource.py#L505 - and kinda tricky for some platforms.

[georgeslabreche]

@roll IteratorChain saved the day! https://commons.apache.org/proper/commons-collections/javadocs/api-2.1.1/org/apache/commons/collections/iterators/IteratorChain.html

Is it OK that we have to explicitly set a base path for the Resource object if the datapackage is using relative file paths?

[roll]

@georgeslabreche Great! (heh at first place I was trying to figure out who has the nickname IteratorChain=)

Base path is even a part of the reference - https://github.com/frictionlessdata/implementations#datapackage

I think in Python/JavaScript now we use current directory as a default base path.

[georgeslabreche]

@roll so current directory as in the directory where in the file system the library is residing or the directory where the app using the library is residing? I'm not 100% convinced this is the best approach, with java at least, because of the potential different possibilities where that library will end up. I'm not yet sure which path is retrieved when using something like System.getProperty("user.dir"); from within the library when it is invoke by some other Main application. Hmm.

[roll]

@georgeslabreche Can't say for Java but the idea that user has a working directory with e.g. datapackage.json. So instead of pushing him to write something like this package = Package('descriptor.json', base_path='.') other libs just use current working directory by default.

[georgeslabreche]

@roll I will test the path retrieval behaviour in Java to make sure I use the proper way of getting the working directory in a predictable manner.