Moving code over to sdg-build

[brockfanning]

Hi @LucyGwilliamAdmin, I'd like to start looking at getting your code over to sdg-build as an "Input". First I want to make sure I understand the basic architecture:

• Step 1 is to convert the DSD into a mapping of codes to human-readable labels
• Step 2 is to parse the SDMX file and use the code-mapping to create a CSV from the human-readable labels

My first attempt at an SDMX import was with SDMX-JSON format, so your SDMX-ML parser will make a nice compliment. Here is an example of what I think we should try to do with your code, by making it into an "Input" in sdg-build: open-sdg/sdg-build#31

[LucyGwilliamAdmin]

Hi @brockfanning, so this would mean that users could upload their SDMX-ML files to the sdg-repo and they'd be converted to SDMX?

Yes, that's correct.

Step 1: since there's a script to convert DSD to CSV, would this involve uploading the DSD to countries data or indicators repo and then mapping created by sdg-build? would that mean that a mapping is created every time the site is built? or could it be done the first time it's uploaded and then that mapping as also saved (say we use UK_DSDv1.2, then the DSD file would be called UK_DSDv1.2.xml and we'd get the sdg-build to save the mapping as UK_mappingv1.2.csv) so that next time its build it can check the version of the both files and if same then skip creating mapping or something? I guess it depends on how long creating the mapping each time takes.

You might have already thought about this, but just some thoughts I had.

[LucyGwilliamAdmin]

Also, I can see the code you linked to for the SDMX-JSON input is in a different format (oop?) to what I've done with the SDMX-ML.

Let me know if you want me to do anything with the SDMX-ML code to make it better to be built into sdg-build - I'm happy to help

[brockfanning]

If we implement your code as an "Input" in sdg-build, this would mean that users could use existing SDMX-ML as a data source for their Open SDG site, instead of the normal approach with CSV files. It might be called "InputSdmxML" for example, and would probably take 2 parameters: the URL of an SDMX-ML file, and the URL of the DSD file. Then it would perform the input/parsing/etc and sdg-build would output everything in the usual JSON needed by Open SDG. So in a nutsell, it would allow using SDMX-ML instead of CSV files.

To the question of caching the mapping of the DSD -- it may be possible to cache this map, but I would recommend against it. It might seem wasteful to repeat the parsing of the DSD on each import, but since this will be automated, we won't notice the extra work. :) And if there is caching then we have to worry about how to recognize changes in the DSD.

So yes the next step I think is to submit your code in the OOP format of an Input class in sdg-build. We could devote a tech call to this process, or have a hangout/hackathon to get it done. I also plan to update my PR for SDMX-JSON to include more documentation and examples.

[brockfanning]

@LucyGwilliamAdmin Could you point me to the code that generated this file?

[LucyGwilliamAdmin]

Hi @brockfanning, that file was generated using ILOSTAT's SMART tool. Some more info about creating the mapping is in this file - I think that is something the country we have to manually.

There is also some more information here

[brockfanning]

Very helpful, thanks @LucyGwilliamAdmin !