Skip to contents

Changelog

blockr 0.0.2.9031

Feature

  • Improved submit feature for blocks. Now submit isn’t added as a class but as a special block attribute. When you design a block, you can pass the submit parameter like so:
new_super_block <- function(submit = NA, ...) {
  fields <- list()
  new_block(
    fields = fields,
    expr = quote(print("test")),
    submit = submit,
    ...,
    class = "my_block"
  )
}

When submit = NA, it will add a submit button but computations are blocked, as clicking on it is required. Internally, once the input$submit is clicked, the submit attribute is set to TRUE. This is useful when the stack is serialized, since this state is kept so that computations can be automatically re-triggered on restore. When submit = TRUE, a button is shown and the result is also computed. When submit = FALSE, no button is shown.

# You can disable the submit button for filter block
serve_stack(new_stack(new_dataset_block(), new_filter_block(columns = "Time", submit = FALSE)))
serve_stack(new_stack(new_dataset_block(), new_filter_block(columns = "Time", submit = NA)))
# Simulate what happens when restoring a serialised stack
serve_stack(new_stack(new_dataset_block(), new_filter_block(columns = "Time", submit = TRUE)))
  • Improved add new block.
  • Added new category to the registry. Now when a block is registered, you may pass a category parameter (which is used by the add block feature to sort blocks):
register_block(
  constructor = new_tail_block,
  name = "tail block",
  description = "return last n rows",
  category = "transform",
  classes = c("tail_block", "transform_block"),
  input = "data.frame",
  output = "data.frame"
)

If not passed, the block will belong to uncategorized blocks (default).

Doc

  • Improved registry and getting started vignettes.
  • Add new case studies vignette to present blockr in various contexts.
  • Refine GitHub readme.

Fixes

  • Fix issue in handle_remove.block: vals$stack wasn’t correctly updated when the last block was removed leading to wrong state.
  • Loading spinner is now correctly hidden when the block visual is updated.
  • Fix #366:
    • We no longer need a reactive poll per result field for the data.
    • The (serialized) field “value” is only the stack name the field refers to (not data).
    • We no longer need a reactive poll for the stack selector drop-down
  • Fix #358.

blockr 0.0.2

Breaking changes

  • Change blocks and fields constructor names to new_*_block and new_*_field. For instance, for the select block, users are now supposed to use new_select_block() and not select_block.
  • Remove data from block constructor:
new_select_block <- function(columns = character(), ...) {
  ...
}

New features

  • New validation functions: validate_field() and validate_block() to check that values are consistent with data. These are used to then propagate any error to the user via JavaScript.
  • Evaluation stops whenever a block isn’t valid and the app should not crash.
  • We can now instantiate block outside the stack with default parameter values (or use the old way with constructors):
# New way
data_blk <- new_dataset_block(selected = "lab", package = "blockr.data")
select_blk <- new_select_block("STUDYID")

stack <- new_stack(data_blk, select_blk)

# Old way
stack <- new_stack(new_data_block, new_select_block)

blockr 0.0.1.9000

  • First prototype:
    • single stacks.
    • data block.
    • filter block.
    • Add/remove block from a stack.