We have some ideas on how to implement data frames supporting
- an unknown number of variables at runtime
- heterogeneous types
- runtime expressions
General syntax with operator|
If you have 3 xvariables (of any data types, including expressions), v1, v2, v3. A "frame" can be obtained with e.g.
auto frame = ("age", v1) | ("weight", v2) | ("gender", v3);the type of frame is parameterized with all the possible types of closure to xvariables that are in the frame, and basically is a map of variants on these types. To augment the frame with a new unevaluated variable e.g. v4 = v1 + v2 * v3, you can do
auto frame2 = frame | ("formula", v1 + v2 * v3);frame2 is an xframe. The variant for the variable types held by frame2 is the union of the variant of frame and the type of v1 + v2 * v3.
Unknown number of variables read from a file without augmenting the type
- In the case where you do something like reading a file. All variables held in the frame have the same type and we don't create a new frame type by adding new variable.
- When you add a large number of variables with
v1 | v2 | v3 | v4, it is evaluated left to right, that is: (((v1 | v2) | v3) | v4). In each case, the left-hand side is an rvalue on an xframe, which can be recycled in the new xframe. In other words, xframe&& | xvariable -> xframe can reuse the input xframe.
We have some ideas on how to implement data frames supporting
General syntax with
operator|If you have 3 xvariables (of any data types, including expressions),
v1,v2,v3. A "frame" can be obtained with e.g.the type of
frameis parameterized with all the possible types of closure to xvariables that are in the frame, and basically is a map of variants on these types. To augment the frame with a new unevaluated variable e.g.v4 = v1 + v2 * v3, you can doframe2is an xframe. The variant for the variable types held byframe2is the union of the variant offrameand the type ofv1 + v2 * v3.Unknown number of variables read from a file without augmenting the type
v1 | v2 | v3 | v4, it is evaluated left to right, that is:(((v1 | v2) | v3) | v4). In each case, the left-hand side is an rvalue on an xframe, which can be recycled in the new xframe. In other words,xframe&& | xvariable -> xframecan reuse the input xframe.