Closed GoogleCodeExporter closed 9 years ago
I have the same issue.
name: name1
email: email@1.com
name: name2
email: email@2.com
name: name3
email: email@3.com
ends up showing just the last data pair as a single row (with name and email
column headings as expected):
name3 email@3.com
Original comment by iae...@gmail.com
on 13 Feb 2012 at 9:06
We could do a better job of this out of the box, but there is a way to work
around the issue.
1. Facet on something which marks the first row of each "record" (e.g. %0 in my
example or "name" in the iaeonh's example)
2. Add column using the expression: row.index
3. Move the resulting column to the beginning
4. Fill down on the column so that all rows in the "record" have the same index
value.
5. Finally, do the key/value transpose and you'll get the desired result.
Original comment by tfmorris
on 18 Sep 2012 at 6:45
p.s. As a breadcrumb for anyone specifically searching for how to do this, my
example file is an EndNote citation file.
Original comment by tfmorris
on 18 Sep 2012 at 6:51
Noting the feature name to find this issue easier: Columnize by key/value
columns
Original comment by thadguidry
on 3 Oct 2012 at 3:16
Attached another use case for this issue to try and resolve it.
The Note field could be choosen with the RECORDNUM column, but that causes only
the last record row to be output. Not choosing the Note Field and only
selecting the Key and Value columns will only extract and columnize against the
first Key value, in this case the "Title" values, but other values in the outer
row processing FOR loop sequence do seem to be populated.
Conversely, removing the RECORDNUM column and starting a Columnize by key/value
columns on only 2 columns, the Key and Value columns, will only return 1 row,
the last record row with all Key fields populated correctly.
Original comment by thadguidry
on 3 Oct 2012 at 4:17
Attachments:
My workaround (or the last piece of it anyway) should work for Thad's example.
Fill Down on the RECORDNUM column, then transpose on Key + Value. I get 86
records with reasonable looking columns.
Original comment by tfmorris
on 3 Oct 2012 at 8:47
Fixed in r2574. I also added support for proper handling of repeated fields
e.g. an EndNote record with multiple author keys. This only works when there
are key, value, and optionally note columns. When there are more columns the
previous/existing algorithm uses a key based on the values of all the remaining
cell values and assumes they they all represent a single record. This allows
the k/v records to be non-continuous but has the downside that it doesn't allow
multiple records with the same non k/v values.
Original comment by tfmorris
on 5 Oct 2012 at 11:39
Original issue reported on code.google.com by
tfmorris
on 9 Feb 2012 at 1:45Attachments: