Improve update_call() function

[bochocki]

Change the name of this function (potentially to call_update()). This function currently needs to be improved for structure and treatment search criteria.

Currently, calls such as:

browse(structure == "age" | structure == "size") fail. This needs to be improved, and tests need to be designed for this function.

[bochocki]

I'm currently working on a general version of call_update() that uses string manipulations. I have a working draft of the code that seems to work in most cases; I just need to clean it up and run some tests before integrating it into the browse() function.

[bochocki]

revised the draft code and implemented it in the browse() function. Here are details on the update:

• changed the function name to call_update, as we discussed.
• added the function code to the file browse.R
• the function can now handle any number of queries to structure and treatment, so that: call_update(substitute(structure == "age" | structure == "size")) returns (structured_type_1 == "age" | structured_type_2 == "age" | structured_type_3 == "age" | structured_type_4 == "age") | (structured_type_1 == "size" | structured_type_2 == "size" | structured_type_3 == "size" | structured_type_4 == "size")
• included support for the following logical operators: &, &&, |, ||, !
• included support for the following comparison operators: ==, !=, <=, >=, >, <, %in%
• tested that output is identical to the current update_call() function:

  identical(call_update(expression(structure == "size" & treatment == "control")),
            update_call(substitute(structure == "size" & treatment == "control")))
  [1] TRUE

• included support for ! A == B syntax
• here's an example of a complex call that tests all features and the resulting output. This is a nonsense example, but demonstrates that the code works (formatted over multiple lines for easier viewing):

  > query <- substitute(
                     treatment == "control" 
                     & !treatment == "burn" 
                     & structure != "size" 
                     | structure =="age" 
                     &!lter_id %in% c(2,3) 
                     && !lter_id >= 5 
                     | lter_id > 4 
                     & lter_id <=200 
                     & lter_id < 42)
  > call_update(query)
  [1] (treatment_type_1 == "control" | treatment_type_2 == "control" | treatment_type_3 == "control") 
       & !(treatment_type_1 == "burn" | treatment_type_2 == "burn" | treatment_type_3 == "burn") 
       & (structured_type_1 != "size" & structured_type_2 != "size" &  structured_type_3 != "size" & structured_type_4 != "size") 
       | (structured_type_1 == "age" | structured_type_2 == "age" | structured_type_3 == "age" | structured_type_4 == "age") 
       & !lter_id %in% c(2, 3) 
       && !lter_id >= 5 
       | lter_id > 4 
       & lter_id <= 200 
       & lter_id < 42

• note that in the code above, the argument structure != "size" gives the output (structured_type_1 != "size" & structured_type_2 != "size" & structured_type_3 != "size" & structured_type_4 != "size"), which strings structure types together with & instead of |. This ensures that no structure types match size.
• checked that the new and old version of browse returned identical results.

  old_browse <- browse(treatment == "fire" & structure == "size")
  new_browse <- browse(treatment == "fire" & structure == "size")
  identical(new_browse,old_browse)
  [1] TRUE

[bochocki]

I suggest we delete update_call.R and close this issue.

[bochocki]

also, for call_update() to work, stringr needs to be in the list of packages that load with popler.

[AldoCompagnoni]

1. I agree on deleting update_call - go ahead if you like.
2. Let's put stringr in section "Depends" of the DESCRIPTION file for now. However, we will have to handle dependencies otherwise, as I was repeatedly suggested not to use "depends".

[bochocki]

deleted update_call.R... I suggest closing this issue once stringr is in the "Depends" section, since that gets into another issue (this one: https://github.com/AldoCompagnoni/popler/issues/23)

[AldoCompagnoni]

stringr is in DESCRIPTION, so I will close this issue!