dynamic underscore selector

[ozgurkalan]

Hi,

How to make a dynamic selection for the underscore ( _ ), especially during aggregation.

sample:

some_var = "quantity"

(t
.aggregate(_.[some_var].sum(), by="year")
)

[gforsyth]

Hey @ozgurkalan -- there are a few ways you can do this:

1. Make use of getattr for grabbing the column name associated with your variable:

[ins] In [23]: var
Out[23]: 'bill_length_mm'

[ins] In [24]: penguins.aggregate(getattr(_, var).sum(), by="year")
Out[24]: 
┏━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━┓
┃ year  ┃ Sum(bill_length_mm) ┃
┡━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━┩
│ int64 │ float64             │
├───────┼─────────────────────┤
│  2007 │              4767.7 │
│  2008 │              4963.7 │
│  2009 │              5289.9 │
└───────┴─────────────────────┘

2. Use a selector

[ins] In [25]: import ibis.selectors as s

[ins] In [26]: var
Out[26]: 'bill_length_mm'

[ins] In [27]: penguins.aggregate(s.across(s.c(var), _.sum()), by="year")
Out[27]: 
┏━━━━━━━┳━━━━━━━━━━━━━━━━┓
┃ year  ┃ bill_length_mm ┃
┡━━━━━━━╇━━━━━━━━━━━━━━━━┩
│ int64 │ float64        │
├───────┼────────────────┤
│  2007 │         4767.7 │
│  2008 │         4963.7 │
│  2009 │         5289.9 │
└───────┴────────────────┘

I'd suggest the selector approach, as it is more flexible and will lend itself more readily to selecting multiple columns in a dynamic way.

[ozgurkalan]

getattr is life saver! Thank you

[jcrist]

Just for completeness - no need to use getattr here. Like tables, columns may be accessed on _ as either _.col_name or _["col_name"].

In [6]: var = "bill_length_mm"

In [7]: penguins.aggregate(_[var].sum(), by="year")
Out[7]: 
┏━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━┓
┃ year  ┃ Sum(bill_length_mm) ┃
┡━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━┩
│ int64 │ float64             │
├───────┼─────────────────────┤
│  2007 │              4767.7 │
│  2008 │              4963.7 │
│  2009 │              5289.9 │
└───────┴─────────────────────┘

[ozgurkalan]

Absoultely! My mind was blown. I happened to put period in between :( Thank you