Create records from data frames (#78)

* Add `from_df()` method to `Registry` * Add temporary record classes with saving * Create a default instance with `connect(slug=NULL)` * Add `delete()` method to Record * Update architecture vignette * Update development vignette Co-authored-by: Robrecht Cannoodt <[email protected]>
laminlabs · Nov 18, 2024 · 804cb06 · 804cb06
1 parent 4e8613e
commit 804cb06
Show file tree

Hide file tree

Showing 19 changed files with 545 additions and 41 deletions.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -2,13 +2,27 @@
 
 ## NEW FUNCTIONALITY
 
-* Add support for more loaders (PR #81).  
+- Add support for more loaders (PR #81).  
   Currently supported: `.csv`, `.h5ad`, `.html`, `.jpg`, `.json`, `.parquet`, `.png`, `.rds`, `.svg`, `.tsv`, `.yaml`.  
-  Planned: `.fcs`, `.h5mu`, `.zarr`.
+- Add a `from_df()` method to the `Registry` class to create new artifacts from data frames (PR #78)
+- Create `TemporaryRecord` classes for new artifacts before they have been saved to the database (PR #78)
+- Add a `delete()` method to the `Record` class (PR #78)
+
+## MAJOR CHANGES
+
+- Running `connect(slug = NULL)` now connects to the default instance that is allowed to create records.
+  The default instance must be changed using the Lamin CLI. (PR #78)
+- User setting are stored in a global option the first time `connect()` is run (PR #78)
+
+## TESTING
+
+- Add a test for creating artifacts from data frames (PR #78).
 
 ## DOCUMENTATION
 
-* Updated installation instructions after **{laminr}** was released on CRAN (PR #74).
+- Updated installation instructions after **{laminr}** was released on CRAN (PR #74).
+- Updated the architecture vignette to include new methods and the new `TemporaryRecord` class (PR #78)
+- Updated the development vignette with new functionality (PR #78)
 
 # laminr v0.1.0
 

diff --git a/NAMESPACE b/NAMESPACE
@@ -24,7 +24,7 @@ importFrom(purrr,map_lgl)
 importFrom(purrr,modify_depth)
 importFrom(purrr,pmap)
 importFrom(purrr,reduce)
-importFrom(purrr,set_names)
 importFrom(purrr,transpose)
 importFrom(purrr,walk)
+importFrom(rlang,set_names)
 importFrom(tools,file_ext)
diff --git a/R/Instance.R b/R/Instance.R
@@ -1,4 +1,4 @@
-create_instance <- function(instance_settings) {
+create_instance <- function(instance_settings, is_default = FALSE) {
   super <- NULL # satisfy linter
 
   api <- InstanceAPI$new(instance_settings = instance_settings)
@@ -46,19 +46,46 @@ create_instance <- function(instance_settings) {
     cloneable = FALSE,
     inherit = Instance,
     public = list(
-      initialize = function(settings, api, schema) {
+      initialize = function(settings, api, schema, is_default, py_lamin) {
         super$initialize(
           settings = settings,
           api = api,
-          schema = schema
+          schema = schema,
+          is_default = is_default,
+          py_lamin = py_lamin
         )
       }
     ),
     active = active
   )
 
+  py_lamin <- NULL
+  if (isTRUE(is_default)) {
+    check_requires("Connecting to Python", "reticulate", type = "warning")
+
+    py_lamin <- tryCatch(
+      reticulate::import("lamindb"),
+      error = function(err) {
+        cli::cli_warn(c(
+          paste(
+            "Failed to connect to the Python {.pkg lamindb} package,",
+            "you will not be able to create records"
+          ),
+          "i" = "See {.run reticulate::py_config()} for more information"
+        ))
+        NULL
+      }
+    )
+  }
+
   # create the instance
-  RichInstance$new(settings = instance_settings, api = api, schema = schema)
+  RichInstance$new(
+    settings = instance_settings,
+    api = api,
+    schema = schema,
+    is_default = is_default,
+    py_lamin = py_lamin
+  )
 }
 
 #' @title Instance
@@ -103,9 +130,13 @@ Instance <- R6::R6Class( # nolint object_name_linter
     #' @param settings The settings for the instance
     #' @param api The API for the instance
     #' @param schema The schema for the instance
-    initialize = function(settings, api, schema) {
+    #' @param is_default Logical, whether this is the default instance
+    #' @param py_lamin A Python `lamindb` module object
+    initialize = function(settings, api, schema, is_default, py_lamin) {
       private$.settings <- settings
       private$.api <- api
+      private$.is_default <- is_default
+      private$.py_lamin <- py_lamin
 
       # create module classes from the schema
       private$.module_classes <- map(
@@ -158,6 +189,12 @@ Instance <- R6::R6Class( # nolint object_name_linter
     get_api = function() {
       private$.api
     },
+    #' @description Get the Python lamindb module
+    #'
+    #' @return Python lamindb module.
+    get_py_lamin = function() {
+      private$.py_lamin
+    },
     #' @description
     #' Print an `Instance`
     #'
@@ -246,9 +283,18 @@ Instance <- R6::R6Class( # nolint object_name_linter
       )
     }
   ),
+  active = list(
+    #' @field is_default (`logical(1)`)\cr
+    #' Whether this is the default instance.
+    is_default = function() {
+      private$.is_default
+    }
+  ),
   private = list(
     .settings = NULL,
     .api = NULL,
-    .module_classes = NULL
+    .module_classes = NULL,
+    .is_default = NULL,
+    .py_lamin = NULL
   )
 )
diff --git a/R/InstanceAPI.R b/R/InstanceAPI.R
@@ -197,6 +197,49 @@ InstanceAPI <- R6::R6Class( # nolint object_name_linter
       private$process_response(response, "get record")
     },
     #' @description
+    #' Delete a record from the instance.
+    delete_record = function(module_name,
+                             registry_name,
+                             id_or_uid,
+                             verbose = FALSE) {
+      user_settings <- .get_user_settings()
+      if (is.null(user_settings$access_token)) {
+        cli::cli_abort(c(
+          "There is no access token for the current user",
+          "i" = "Run {.code lamin login} and reconnect to the database in a new R session"
+        ))
+      }
+
+      url <- paste0(
+        private$.instance_settings$api_url,
+        "/instances/",
+        private$.instance_settings$id,
+        "/modules/",
+        module_name,
+        "/",
+        registry_name,
+        "/",
+        id_or_uid,
+        "?schema_id=",
+        private$.instance_settings$schema_id
+      )
+
+      if (verbose) {
+        cli_inform("URL: {url}")
+      }
+
+      response <- httr::DELETE(
+        url,
+        httr::add_headers(
+          accept = "application/json",
+          `Content-Type` = "application/json",
+          Authorization = paste("Bearer", user_settings$access_token)
+        )
+      )
+
+      private$process_response(response, "delete record")
+    },
+    #' @description
     #' Print an `API`
     #'
     #' @param style Logical, whether the output is styled using ANSI codes

diff --git a/R/Record.R b/R/Record.R
@@ -45,6 +45,78 @@ create_record_class <- function(instance, registry, api) {
   RichRecordClass
 }
 
+#' Create a temporary record class
+#'
+#' @param record_class A generator for a standard record class
+#'
+#' @details
+#' The classes generated by this function inherit from a standard record class
+#' and represent the situation where a new record has being created using Python
+#' `lamindb` but has not yet been saved to the database. It should behave the
+#' same as the standard class but indicate to the user that it has not yet been
+#' saved. Saving is performed using the Python record object stored in the R
+#' temporary record. After saving, the data in the object is replaced with that
+#' in the database, indications that it has not been saved are removed and
+#' further saving is prevented. In examples etc. the `$save()` should usually be
+#' called immediately to avoid users seeing the temporary record in it's unsaved
+#' state.
+#'
+#' @return The temporary record class R6 generator
+#' @noRd
+create_temporary_record_class <- function(record_class) {
+  super <- NULL # Satisfy checks
+  self <- NULL # Satisfy checks
+  private <- NULL # Satisfy checks
+
+  R6::R6Class(
+    paste0("Temporary", record_class$classname),
+    cloneable = FALSE,
+    inherit = record_class,
+    public = list(
+      initialize = function(py_record, data) {
+        private$.record_class <- record_class
+        private$.py_record <- py_record
+
+        super$initialize(data)
+      },
+      save = function() {
+        if (isTRUE(private$.saved)) {
+          cli::cli_abort("This record has already been saved to the database")
+        }
+
+        private$.py_record$save()
+
+        # Replace temporary data with data saved to the database
+        private$.data <- private$.api$get_record(
+          module_name = private$.registry$module$name,
+          registry_name = private$.registry$name,
+          id_or_uid = self$uid
+        )
+
+        private$.saved <- TRUE
+      },
+      print = function(style = TRUE) {
+        if (isFALSE(private$.saved)) {
+          cli::cat_line(paste(
+            cli::bg_red(cli::col_black("TEMPORARY")),
+            cli::format_message(paste(
+              "This record has not been saved to the database.",
+              "Save it using {.code <object>$save()}."
+            ))
+          ))
+        }
+
+        super$print()
+      }
+    ),
+    private = list(
+      .record_class = NULL,
+      .py_record = NULL,
+      .saved = FALSE
+    )
+  )
+}
+
 #' @title Record
 #'
 #' @description
@@ -98,6 +170,22 @@ Record <- R6::R6Class( # nolint object_name_linter
       }
     },
     #' @description
+    #' Delete a `Record`
+    #'
+    #' @param verbose Whether to print details of the API call
+    #'
+    #' @return `TRUE` invisibly if the deletion is successful
+    delete = function(verbose = FALSE) {
+      response <- private$.api$delete_record(
+        module_name = private$.registry$module$name,
+        registry_name = private$.registry$name,
+        id_or_uid = self$uid,
+        verbose = verbose
+      )
+
+      invisible(TRUE)
+    },
+    #' @description
     #' Print a `Record`
     #'
     #' @param style Logical, whether the output is styled using ANSI codes
@@ -120,12 +208,28 @@ Record <- R6::R6Class( # nolint object_name_linter
         "key"
       )
 
-      record_fields <- private$.api$get_record(
-        module_name = private$.registry$module$name,
-        registry_name = private$.registry$name,
-        id_or_uid = private$.data[["uid"]],
-        include_foreign_keys = TRUE
-      )
+      expected_fields <- private$.registry$get_fields() |>
+        discard(~ is.null(.x$column_name)) |>
+        map_chr("column_name")
+
+      record_fields <- map(names(expected_fields), function(.field) {
+        value <- tryCatch(
+          self[[.field]],
+          error = function(err) {
+            if (!grepl("status code 404", conditionMessage(err))) {
+              cli::cli_abort(conditionMessage(err))
+            }
+            NULL
+          }
+        )
+
+        if (inherits(value, "Record")) {
+          value <- value$id
+        }
+
+        value
+      }) |>
+        set_names(expected_fields)
 
       # Get the important fields that are in the record
       important_fields <- intersect(important_fields, names(record_fields))