Hello again,

As previously mentioned, I'm trying to build an extensible and data type agnostic extension to the CSV functionality and I have the following questions and ideas and things I've tried:

1) Can a primitive type be assigned to Data? I want to pass an array of primitives (the new column or row) but that array could be either ints or decs and I don't know which at compile time. So how do I set up the function signature?

2) I thought the problem solved by the ADT library, like this: "void add(store Data item)" - this is from adt.List. However, it doesn't compile when I try to use store.

3) Do I wrap primitive types in Data? Like String?

4) Do I provide a bunch of the functions for different primitive data types?

5) Do I do it with optional parameters and set only the one I'm interested in?

6) Do I serialise Data and then deserialise it based on a Type provided in the function?

I went with the idea of wrapping primitive data types in Data, however, this causes another issue with assignment between variables with different sizes. I'll create a separate post for this issue.

Again, I'm open to suggestions!

Many thanks,

Sava

HI Sava,

You can't automatically assign a primitive type to a Data. The store qualifier is for a different purpose -- it's to indicate whether or not a function's implementation intends to keep a given reference type in its global state. This is a different topic, but this is necessary for stateful side effect analysis for sound runtime adaptation support.

If I understand correctly you probably are looking to design an interface something like this:

interface CSVColumn {
   CSVColumn(char path[])
   void addColumnInt(char heading[], int values[])
   void addColumnDec(char heading[], dec values[])
   }

I think the above would be sufficient since you only really need to support int and dec types here. Your idea of using optional parameters would also work well, though! i guess there you'd have list one addColumn() function which perhaps always takes a heading, and optionally takes an int or dec array?

I think this is a scenario in which there isn't an existing precedent in the standard library, so it's really up to you to choose the design you think is best :-)

Barry

Hello,

Here is the latest version of my work:

https://github.com/savakazakov/CSVUtilExtension

I would love to get some opinions on matter from more experienced Dana users.

Many thanks,Sava

Hi,

I just had a chance to look at this. I think you have the start of something really quite interesting here!

I would simplify it a bit to make it easy for regular int[] and dec[] arrays, plus possibly char[]-based columns (via String[]).

I also thought it might be a really useful addition to have a function which reads in the entire CSV file into a Data`. People could then combine this quite easily with the stats.chart stuff to make graphs.

Thanks for sharing!

Hi Sava :-)

I totally agree with the above comments, and the way this is implemented is really cool :-)

With maybe a bit more expansion it would be great to see something like this added to the standard library.

jess

Hello,

Thanks for the suggestions. I noted those and some give to me by Barry Porter and I'll fix them in the coming days. Moreover, I have some ideas for a slightly different function signatures and I'd really appreciate your input on what you think fits best given the use case.

Once again, thank you very much!

Sava