# Help generating an Expression

**URL:** <https://discourse.julialang.org/t/help-generating-an-expression/47283>\
**Category:** General Usage\
**Tags:** metaprogramming, dataframes\
**Created:** [September 25, 2020, 7:25pm UTC](https://discourse.julialang.org/t/help-generating-an-expression/47283 "2020-09-25T19:25:26Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![mthelm85](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/mthelm85/32/224164_2.png) [@mthelm85](https://discourse.julialang.org/u/mthelm85)\
**Post date:** [September 25, 2020, 7:25pm UTC](https://discourse.julialang.org/t/help-generating-an-expression/47283/1 "2020-09-25T19:25:26Z")

</div>

I have some wide-formatted data that I need to group and summarize. A MWE is below:

```nohighlight
using DataFramesMeta

df = DataFrame(
    id = repeat(1:5, inner=3),
    a_1 = rand(15),
    a_2 = rand(15),
    a_3 = rand(15),
    b_1 = rand(15),
    b_2 = rand(15),
    b_3 = rand(15)
)

summary = @linq df |>
    by(
        :id,
        a_1 = sum(:a_1),
        a_2 = sum(:a_2),
        a_3 = sum(:a_3),
        b_1 = sum(:b_1),
        b_2 = sum(:b_2),
        b_3 = sum(:b_3),
    )

julia> summary
5×7 DataFrame
│ Row │ id │ a_1 │ a_2 │ a_3 │ b_1 │ b_2 │ b_3 │
│ │ Int64 │ Float64 │ Float64 │ Float64 │ Float64 │ Float64 │ Float64 │
├─────┼───────┼─────────┼──────────┼──────────┼──────────┼─────────┼─────────┤
│ 1 │ 1 │ 2.34505 │ 0.881786 │ 0.969993 │ 0.635778 │ 1.69327 │ 1.22705 │
│ 2 │ 2 │ 1.24684 │ 0.488665 │ 1.65159 │ 1.85681 │ 1.63358 │ 1.05604 │
│ 3 │ 3 │ 2.05449 │ 0.928822 │ 1.53021 │ 1.66948 │ 1.8532 │ 1.69901 │
│ 4 │ 4 │ 1.77823 │ 1.19143 │ 1.36264 │ 1.13891 │ 1.90184 │ 1.46904 │
│ 5 │ 5 │ 1.00121 │ 1.91585 │ 1.11492 │ 1.69419 │ 1.58653 │ 1.98337 │

```

This is fine but in my real example I have hundreds of columns so I don’t want to write out

```julia
    by(
        :id,
        a_1 = sum(:a_1),
        a_2 = sum(:a_2),
        a_3 = sum(:a_3),
        b_1 = sum(:b_1),
        b_2 = sum(:b_2),
        b_3 = sum(:b_3),
    )

```

this portion of the code. The column names follow a pattern so I was hoping to be able to generate an `Expr` that I could `eval` inside of `by` but I cannot figure out how to do this.

Something like:

```julia
julia> for l in (:a,:b), i in 1:3
           println("$(l)_$i = sum(:$(l)_$i),")
       end
a_1 = sum(:a_1),
a_2 = sum(:a_2),
a_3 = sum(:a_3),
b_1 = sum(:b_1),
b_2 = sum(:b_2),
b_3 = sum(:b_3),

```

but creating an `Expr` that I can inject into the `by` function above and `eval`.

Is such a thing possible or is there a better way to do this? Or am I stuck writing it all out?

---

<div class="post-metadata">

**Author:** ![pdeffebach](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/pdeffebach/32/10320_2.png) [@pdeffebach](https://discourse.julialang.org/u/pdeffebach)\
**Post date:** [September 25, 2020, 7:29pm UTC](https://discourse.julialang.org/t/help-generating-an-expression/47283/2 "2020-09-25T19:29:53Z")

</div>

Don’t use dataframesmeta for this, just use dataframes!

```julia
combine(groupby(df, :id), names(df, r"^a_") .=> sum, names(df, r"^b_") .=> sum)

```

---

<div class="post-metadata">

**Author:** ![mthelm85](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/mthelm85/32/224164_2.png) [@mthelm85](https://discourse.julialang.org/u/mthelm85)\
**Post date:** [September 25, 2020, 7:31pm UTC](https://discourse.julialang.org/t/help-generating-an-expression/47283/3 "2020-09-25T19:31:51Z")

</div>

> [@pdeffebach](#):
>
> Don’t use dataframesmeta for this, just use dataframes!

I’ve been using DataFramesMeta for so long now that I forget how good regular ol’ DataFrames is 😉

---

<div class="post-metadata">

**Author:** ![pdeffebach](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/pdeffebach/32/10320_2.png) [@pdeffebach](https://discourse.julialang.org/u/pdeffebach)\
**Post date:** [September 25, 2020, 7:34pm UTC](https://discourse.julialang.org/t/help-generating-an-expression/47283/4 "2020-09-25T19:34:09Z")

</div>

Well, this is (relatively) new functionality.

DataFramesMeta master right now uses `transform`, `combine` etc. as the back-end for `@transform`, `@by` etc.

So it should be prototype something in DataFramesMeta and then translate the expression to something without any literals and write it in DataFrames.