I’ve thought about this for a recent project but didn’t implement it, because I decided it wasn’t worth the code for SQL vanity purpose.

In your reducer function, you could case on the accumulator condition and, if nil, produce a query without the and. The benefit of using “true and” is that your reducer doesn’t need to care what the accumulator is.

I haven’t tried it but I think this would work.

Thank you @sb8244! It’s simpler than I thought, just made it

Before:

defp parse_item({compare_type, value}, acc, field_name) do
  case compare_type do
    :eq -> dynamic([p], ^acc and field(p, ^field_name) == ^value)
    :gt -> dynamic([p], ^acc and field(p, ^field_name) > ^value)
    :lt -> dynamic([p], ^acc and field(p, ^field_name) < ^value)
    :gte -> dynamic([p], ^acc and field(p, ^field_name) >= ^value)
    :lte -> dynamic([p], ^acc and field(p, ^field_name) <= ^value)
    _ -> acc
  end
end

After:

defp parse_item({compare_type, value}, acc, field_name) do
  case compare_type do
    :eq -> join_exp(acc, dynamic([p], field(p, ^field_name) == ^value))
    :gt -> join_exp(acc, dynamic([p], field(p, ^field_name) > ^value))
    :lt -> join_exp(acc, dynamic([p], field(p, ^field_name) < ^value))
    :gte -> join_exp(acc, dynamic([p], field(p, ^field_name) >= ^value))
    :lte -> join_exp(acc, dynamic([p], field(p, ^field_name) <= ^value))
    _ -> acc
  end
end

defp join_exp(acc, exp) do
  case inspect(acc) do
    "dynamic([], true)" ->
      exp

    _ ->
      dynamic(^acc and ^exp)
  end
end

Relying on inspect's output is awful but sometimes a fact of life.