assign side_effects and occurrence as attributes of @code

[m7pr]

Part of #216

Changes:

• moved dependency extraction from get_code_dependency to eval_code
• removed extract_code_graph
• extended documentation of qenv with 1 new attributes: dependency, occurrence
• merged side_effects and occurrence inside eval_code as they were previously joined in extract_code_graph anyway
• created tests for qenv() |> eval_code |> get_code_attr("dependency")
• changed extract_side_effects and extract_occurrence so they work on an element, and they don't use lapply

[m7pr]

Hey @gogonzo - if you have some spare time, this PR is ready for the first review.
It aims for transition of code_graph calculations from get_code_dependency into single calls in eval_code

[gogonzo]

extract_calls while calls are already split?

[m7pr]

extract_occurrence assume that pd has a specific order of rows, that gets changed when you apply extract_calls(pd) on pd.

I can move call_pd <- extract_calls(pd)[[1]] inside extract_occurrence. Or is it ok to keep in eval_code?

[gogonzo]

Let me ask another question - why to reorder pd?

[m7pr]

extract_occurrence assumes object names (SYMBOL) appear after <- (LEFT_ASSIGN/ASSIGN).

extract_calls reorders pd in this way.

code <- "a<-1"
parsed_code <- parse(text = code, keep.source = TRUE)
pd <- utils::getParseData(parsed_code)
> pd
  line1 col1 line2 col2 id parent       token terminal text
7     1    1     1    4  7      0        expr    FALSE     
1     1    1     1    1  1      3      SYMBOL     TRUE    a
3     1    1     1    1  3      7        expr    FALSE     
2     1    2     1    3  2      7 LEFT_ASSIGN     TRUE   <-
4     1    4     1    4  4      5   NUM_CONST     TRUE    1
5     1    4     1    4  5      7        expr    FALSE     
> extract_calls(pd)
  line1 col1 line2 col2 id parent       token terminal text
7     1    1     1    4  7      0        expr    FALSE     
3     1    1     1    1  3      7        expr    FALSE     
2     1    2     1    3  2      7 LEFT_ASSIGN     TRUE   <-
5     1    4     1    4  5      7        expr    FALSE     
1     1    1     1    1  1      3      SYMBOL     TRUE    a
4     1    4     1    4  4      5   NUM_CONST     TRUE    1

If we don't want to use extract_calls we need to revisit and rewrite extract_occurrence. I am not sure we will invent rules to figure out dependencies for a non-sorted pd.

So right now this is a matter of whether we

• use extract_calls on pd and before we apply pd inside extract_occurrence
• or whether we put extract_calls inside extract_occurrence
• or whether we rewrite extract_occurrence.

extract_occurrence is a pretty big beast : P

[m7pr]

Maybe we can simplify extract_calls part so that it only reorders

[m7pr]

I just played with what parts of extract_calls are needed so that extract_occurrence works and below are the needed parts:

# reordering basen on parent-children relation
parent_ids <- pd[pd$parent == 0 & (pd$token !e= "COMMENT" | grepl("@linksto", pd$text, fixed = TRUE)), "id"]
pd_order <- do.call(rbind, lapply(parent_ids, function(parent) rbind(pd[pd$id == parent, ], get_children(pd, parent))))
# filtering or edge cases
if (!is.null(pd_order) && !(nrow(pd_order) == 1 && call$token == "';'")) {
    # fixing assignment arrows
    pd_order <- fix_arrows(list(pd_order))[[1]]
    attr(current_code, "dependency") <- c(extract_side_effects(pd_order), extract_occurrence(pd_order))
}

So this is basically all of extract_calls without fix_shifted_comments that is skipped if there is only 1 call.

If extract_calls would be renamed to reorder_and_clean_calls then I suppose usage of this function is totally justified in here. We can always put this part in extract_occurrence but it's gonna be repeated in extract_calls and extract_occurrence

[m7pr]

Hey @gogonzo I also needed to change the way we create teal.data::teal_data()
a0d4890b7816b614ba353be8db2175ccfb6c39ed
so that it's @code field corresponds to the changes made in here.

There is a bit code from teal.code for unexported functions. Should we export something in teal.code, so that it could be reused in teal.data?

[gogonzo]

This loop looks nice now. extract_dependency is a good move 👍

[gogonzo]

All good here 👍 Please see the comment below