Nested JSON and never end Records

[ElliotP]

Afternoon,

I have a json file (https://1drv.ms/u/s!At8Q-ZbRnAj8hkJLL1cyU4t_hoHC) which has many nested arrays and I'm unsure of how to extract all of the records into powerbi.

I've watch the guyinacube video (incredibly helpful), but I get to a point where I can't expand my columns any further yet I need the data inside that record (it's actually the data I want).

First step: https://gyazo.com/86e411d5de9d7f2f52813f6f66cb3bf9

Convert to table, easy it converts out to this.

Second Step: https://gyazo.com/546206f9a6b0459a4a09b2c865c3d902

Expand out, again comfortable. Expand again.

Third Step (issue): https://gyazo.com/ffe36ce86fa91f4ae0a435c524d026e5

The column on the far right is the colum which contains the data that i want, but I'm unable to retrieve the data from that column without individually clicking on the links.

Any ideas as to how I would be to produce rows or tables from this?

[astoz]

Hello, this post is really useful thanks, It helped me solving an issue I had

I'd like to check my understanding of this piece of code in the solution:

if _ is record then {_} else _}

Am I right in assuming that here it says if current value is record then store it in an object else go to the next value? I'm asking cause I'm not yet entirely clear on what _ (underscore) means in Power Query

[ImkeF]

Yes, normally you would create a function that opens and transforms one (sample) record and apply this in for each cell by calling it from a new column.

But this would only work, if all items have the same structure. This doesn't seem to be the case for your data here, as there are records and lists in them. So you would at least need two different functions and apply them using a conditional statement in the new column.

Or do you know in advance that some of these rows will not be needed and can then filter them out before the expansion?

[ElliotP]

I know all of the records will have the same structure (columns names and data types). Would it be best to somehow create their own table and expand if possible somehow there?

[ImkeF]

Hi @ElliotP,

yes, it's very unlikely that it makes sense to expand 2 different table structures into one column.

So you can add a column to your current query that you can use as a filter wit this:  

Value.Is([Column1.JournalLines.JournalLine], type table)

This checks if the type of the value is table and returns true or false.

Then you reference this table once and set a filter on true, and in a second query a filter on false.

Check out @hugoberry's function, it is very promising.

@hugoberry: Kudos to your function repository! Yesterday I've discovered your Matrix-multiplication - that's a real beauty! How do you do the indentation in your code? Manually or do you use a program?

[hugoberry]

Hey @ImkeF thanks for mentioning that 😉

More posts/queries are coming. I do my identation manually, because I still can't make my visual studio code extension to work properlly. Once I get one in a more via ble stage I'll give you a heads up 😉

[ImkeF]

Thx @hugoberry can't wait (for both) 😉

[ElliotP]

@hugoberryYou are a superstar. Thank you so much, I can't thank you enough, this is amazing.

@ImkeFThank you for your help, I really the time and the ideas. I liked our ideas of breaking the tables apart and linking.

[hugoberry]

Hi @ElliotP, my initial reply was quite generic. Only now I had a chance to look at your JSON.

I've managed to drill down to the data that you were after. If I understand right the format of your data, at the step where the column becomes either a list or a record you have to apply a transofrmation of cell contents and cast them into a list, and then use standard expand procedures to expand the list values in the table.

Here is the step that will help:

= Table.TransformColumns(step, {"Column1.JournalLines.JournalLine", each if _ is record then {_}  else _})

[hugoberry · Accepted solution]

Here is the final result that you can follow

let
    json= Json.Document(File.Contents("D:\Downloads\Xero Datapowerbiforum")),
    json_tab = Table.FromList(json, Splitter.SplitByNothing()),
    expand_1 = Table.ExpandRecordColumn(json_tab, "Column1", {"JournalID", "JournalDate", "JournalNumber", "CreatedDateUTC", "SourceID", "SourceType", "JournalLines"}),
    expand_2 = Table.ExpandRecordColumn(expand_1, "JournalLines", {"JournalLine"}),
    journal_line_transform = Table.TransformColumns(expand_2, {"JournalLine", each if _ is record then {_}  else _}),
    expand_3 = Table.ExpandListColumn(journal_line_transform, "JournalLine"),
    expand_4 = Table.ExpandRecordColumn(expand_3, "JournalLine", {"JournalLineID", "AccountID", "AccountCode", "AccountType", "AccountName", "Description", "NetAmount", "GrossAmount", "TaxAmount", "TaxType", "TaxName"}, {"JournalLineID", "AccountID", "AccountCode", "AccountType", "AccountName", "Description", "NetAmount", "GrossAmount", "TaxAmount", "TaxType", "TaxName"})
in
    expand_4

[hugoberry]

Hi @ElliotP I've tried to come up with an algortihm to do the extraction automatically for a JSON object. See maybe you can use this function for your particular JSON

https://gist.github.com/Hugoberry/4ad49f4301edf47fffe2ef06aed61513