The ultimate Fabric, Power BI, SQL, and AI community-led learning event. Save €200 with code FABCOMM.
Get registeredCompete to become Power BI Data Viz World Champion! First round ends August 18th. Get started.
I want to condionally fill down multiple columns (approximatively 200 columns) as below :
When the value's column is "E" => begin to fill down, and stop when the value "H" is reached.
Example of 2 columns input data :
Column1 | Column2 |
A | W |
B | X |
C | Y |
D | Z |
null | null |
null | E |
null | null |
E | F |
null | null |
null | null |
null | null |
F | G |
G | H |
null | null |
null | null |
H | null |
null | null |
Desired output :
Column1 | Column2 |
A | W |
B | X |
C | Y |
D | Z |
null | null |
null | E |
null | E |
E | F |
E | F |
E | F |
E | F |
F | G |
G | H |
G | null |
G | null |
H | null |
null | null |
Thanks
Solved! Go to Solution.
let
Source = your_table,
to_cols = Table.ToColumns(Source),
g = (lst) =>
[st = [E = true, H = false],
gen = List.Generate(
() => [i = 0, c = lst{0}, fd = c = "E"],
(x) => x[i] < List.Count(lst),
(x) =>
[
i = x[i] + 1,
c = if x[fd] then lst{i} ?? x[c] else lst{i},
fd = Record.FieldOrDefault(st, c ?? "", x[fd])
],
(x) => x[c]
)][gen],
tra = List.Transform(to_cols, g),
to_tbl = Table.FromColumns(tra)
in
to_tbl
Hi @dhendus, different approach here. Test speed and let me know:
Result:
let
Source = Table.FromRows(Json.Document(Binary.Decompress(Binary.FromText("i45WclTSUQpXitWJVnICsiLALGcgKxLMcgGyosAsIANGu6IIuAIZbuhKUGg3IMMdzHIHsjywqvFAF/DEYponioAXkoyXUmwsAA==", BinaryEncoding.Base64), Compression.Deflate)), let _t = ((type nullable text) meta [Serialized.Text = true]) in type table [Column1 = _t, Column2 = _t]),
// You can probably delete this step.
ReplaceBlankToNull = Table.TransformColumns(Source, {}, each if _ = "" then null else _),
Transform = List.Accumulate(
List.Buffer(Table.ColumnNames(ReplaceBlankToNull)),
#table({"_Temp_"}, {{null}}),
(state, current)=> Table.AddColumn(state, current, each
[ a = Table.SelectColumns(ReplaceBlankToNull, current),
b = Table.Group(a, {current}, {{"All", each _, type table}}, GroupKind.Local, (w,u)=> Byte.From( (Record.Field(w, current) <> "E" and Record.Field(u, current) = "E") or (Record.Field(w, current) = "E" and Record.Field(u, current) = "H") ) ),
c = Table.AddColumn(b, "Filled", (x)=> if Record.Field(x, current) = "E" then Table.FillDown(x[All], {current}) else x[All], type table),
d = Table.Combine(c[Filled]),
e = Table.Column(d, current)
][e], type list)
),
RemovedColumns = Table.RemoveColumns(Transform,{"_Temp_"}),
ToTable = Table.FromColumns(Table.ToRows(RemovedColumns){0})
in
ToTable
let
Source = your_table,
to_cols = Table.ToColumns(Source),
g = (lst) =>
[st = [E = true, H = false],
gen = List.Generate(
() => [i = 0, c = lst{0}, fd = c = "E"],
(x) => x[i] < List.Count(lst),
(x) =>
[
i = x[i] + 1,
c = if x[fd] then lst{i} ?? x[c] else lst{i},
fd = Record.FieldOrDefault(st, c ?? "", x[fd])
],
(x) => x[c]
)][gen],
tra = List.Transform(to_cols, g),
to_tbl = Table.FromColumns(tra)
in
to_tbl
That's what I'm talking about, thank you.
Unfortunately, I have memory issues when loading data, in step "Table.FromColumns(tra)" (can not load my data, I've been waiting for an hour and yet not finishing)
can you suggest something else for this step ?
try to buffer to_cols step:
to_cols = List.Buffer(Table.ToColumns(Source))
Thanks, it dropped to 10 min 🙂