# From R-universe
install.packages("automerge", repos = "https://posit-dev.r-universe.dev")

# From GitHub
pak::pak("posit-dev/automerge-r")

library(automerge)

# Create
doc <- am_create() # New document
doc <- am_create("fd3ca50687f13477f1f9ea2b216b958a") # With custom actor ID

# Load/Save
bytes <- am_save(doc) # Save to bytes
doc <- am_load(bytes) # Load from bytes

# Actor ID (document-bound)
actor <- am_get_actor(doc) # Get actor ID (raw bytes)
actor_hex <- am_get_actor_hex(doc) # Get actor ID as hex string
am_set_actor(doc, actor_hex) # Set actor ID (raw, hex, or NULL)

# Fork/Merge/Clone
doc2 <- am_fork(doc) # Create independent copy
doc3 <- am_clone(doc) # Deep copy (independent document)
am_merge(doc, doc2) # Merge doc2 into doc1

# Compare documents
am_equal(doc, doc2) # TRUE if same state

# Transactions
am_commit(doc, "message") # Commit changes
am_rollback(doc) # Cancel pending changes
am_pending_ops(doc) # Number of uncommitted operations
am_commit_empty(doc, "checkpoint") # Create empty change (bookkeeping)

# Incremental save/load
inc <- am_save_incremental(doc) # Only changes since last save
am_load_incremental(doc, inc) # Apply incremental changes

# Cleanup
am_close(doc) # Explicitly free resources (optional)

# S3 operators
doc[["key"]] <- "value" # Set value
value <- doc[["key"]] # Get value
doc$key <- value # Alternative syntax
value <- doc$key # Alternative syntax

# Functional API
am_delete(doc, AM_ROOT, "key") # Delete key
am_put(doc, AM_ROOT, "key", value) # Set value
value <- am_get(doc, AM_ROOT, "key") # Get value

# Introspection
keys <- names(doc) # Get all keys
keys <- am_keys(doc, AM_ROOT) # Functional version
values <- am_values(doc, AM_ROOT) # Get all values
n <- length(doc) # Number of keys
n <- am_length(doc, AM_ROOT) # Functional version
items <- am_items(doc, AM_ROOT) # All key-value entries

# All conflicting values for a key (after concurrent edits)
conflicts <- am_map_get_all(doc, AM_ROOT, "key")

# Range of entries by key (alphabetical)
entries <- am_map_range(doc, AM_ROOT, "a", "m")

# Automatic recursive conversion (recommended)
am_put(
  doc,
  AM_ROOT,
  "user",
  list(
    name = "Alice",
    age = 30L,
    address = list(city = "NYC", zip = 10001L)
  )
)

# Path-based access (simple for deep structures)
am_put_path(doc, c("user", "address", "city"), "Boston")
city <- am_get_path(doc, c("user", "address", "city"))
am_delete_path(doc, c("user", "address"))

# Manual access
user_obj <- am_get(doc, AM_ROOT, "user")
am_put(doc, user_obj, "name", "Bob")
name <- am_get(doc, user_obj, "name")

# Create list
am_put(doc, AM_ROOT, "items", AM_OBJ_TYPE_LIST)
items <- am_get(doc, AM_ROOT, "items")

# Operations (1-based indexing)
am_insert(doc, items, 1, "first") # Insert at index 1
am_put(doc, items, 1, "FIRST") # Replace at index 1
am_put(doc, items, "end", "append") # Append to end
value <- am_get(doc, items, 1) # Get index 1
am_delete(doc, items, 1) # Delete index 1

# Introspection
n <- am_length(doc, items) # List length
items_entries <- am_items(doc, items) # All index-value entries

# All conflicting values at index (after concurrent edits)
conflicts <- am_list_get_all(doc, items, 1)

# Subrange (1-based indexing)
sub <- am_list_range(doc, items, 2, 4)

# Text objects use 0-based inter-character positions
# For the text "Hello":
#  H e l l o
# 0 1 2 3 4 5  <- positions (0-based, between characters)

# Create text object
am_put(doc, AM_ROOT, "content", am_text("Hello"))
text_obj <- am_get(doc, AM_ROOT, "content")

# Operations
am_text_splice(text_obj, 5, 0, " World") # Insert at position 5
content <- am_text_content(text_obj) # Get full text

# Update text (ideal for collaborative editing)
old_text <- am_text_content(text_obj)
am_text_update(text_obj, old_text, "Hello Universe") # Computes and applies diff

# Cursors (stable positions)
cursor <- am_cursor(text_obj, 5) # Create at position 5
pos <- am_cursor_position(cursor) # Get current position
cursor <- am_cursor(text_obj, 5, heads) # Create at historical state
pos <- am_cursor_position(cursor, heads) # Position at historical state

# Cursor serialization (persistence across sessions)
bytes <- am_cursor_to_bytes(cursor) # Serialize to raw
cursor <- am_cursor_from_bytes(bytes, text_obj) # Restore from raw
str <- am_cursor_to_string(cursor) # Serialize to string
cursor <- am_cursor_from_string(str, text_obj) # Restore from string
am_cursor_equal(cursor1, cursor2) # Compare cursors

# Marks (formatting)
am_mark(text_obj, 0, 5, "bold", TRUE, expand = "none")
marks <- am_marks(text_obj) # Get all marks
marks <- am_marks(text_obj, heads) # Marks at historical state
marks_at <- am_marks_at(text_obj, 2) # Marks at position 2
marks_at <- am_marks_at(text_obj, 2, heads) # At position, historical state
am_mark_clear(text_obj, 0, 5, "bold") # Remove a mark from range

# NULL
am_put(doc, AM_ROOT, "null", NULL)

# Boolean
am_put(doc, AM_ROOT, "bool", TRUE)

# Integer
am_put(doc, AM_ROOT, "int", 42L)

# Double
am_put(doc, AM_ROOT, "float", 3.14)

# String
am_put(doc, AM_ROOT, "str", "text")

# Raw bytes
am_put(doc, AM_ROOT, "bytes", raw(10))

# Timestamp
am_put(doc, AM_ROOT, "time", Sys.time())

# Counter
am_put(doc, AM_ROOT, "score", am_counter(0))
am_counter_increment(doc, AM_ROOT, "score", 10)
value <- am_get(doc, AM_ROOT, "score")

# Unsigned 64-bit integer (for cross-platform interop)
am_put(doc, AM_ROOT, "id", am_uint64(12345))

# Explicit type constructors
am_put(doc, AM_ROOT, "items", am_list()) # Empty list
am_put(doc, AM_ROOT, "config", am_map()) # Empty map
am_put(doc, AM_ROOT, "notes", am_text()) # Text object

# Bidirectional sync (auto-converge)
# Documents are modified in place
rounds <- am_sync(doc1, doc2)
cat("Converged in", rounds, "rounds\n")

# One-way merge
am_merge(doc1, doc2) # Merge doc2 into doc1

# Create sync state
sync_state <- am_sync_state()

# Encode/decode messages
msg <- am_sync_encode(doc, sync_state)
am_sync_decode(doc, sync_state, msg)

# Serialize/restore sync state (for persistent connections)
sync_bytes <- am_sync_state_encode(sync_state)
sync_state <- am_sync_state_decode(sync_bytes)

# Manual sync loop
repeat {
  msg1 <- am_sync_encode(doc1, sync1)
  if (is.null(msg1)) {
    break
  }
  am_sync_decode(doc2, sync2, msg1)

  msg2 <- am_sync_encode(doc2, sync2)
  if (is.null(msg2)) {
    break
  }
  am_sync_decode(doc1, sync1, msg2)
}

# Get heads (fingerprint of current state)
heads <- am_get_heads(doc)

# Get changes (returns am_change objects)
changes <- am_get_changes(doc, NULL)  # All changes
changes <- am_get_changes(doc, heads) # Changes since heads (i.e. none)

# Typical pattern: remember a checkpoint, then get new changes later
checkpoint <- am_get_heads(doc)
am_put(doc, AM_ROOT, "new_key", "new_value")
am_commit(doc, "Add new_key")
new_changes <- am_get_changes(doc, checkpoint) # Only the new commit

# Apply changes
am_apply_changes(doc, changes)

# Full history (returns am_change objects)
history <- am_get_changes(doc)

# Change introspection
change <- history[[1]]
hash <- am_change_hash(change) # 32-byte hash
msg <- am_change_message(change) # Commit message or NULL
time <- am_change_time(change) # POSIXct timestamp
actor <- am_change_actor_id(change) # Actor ID (raw bytes)
seq <- am_change_seq(change) # Sequence number (double)
deps <- am_change_deps(change) # List of parent hashes
size <- am_change_size(change) # Number of operations
bytes <- am_change_to_bytes(change) # Serialize to raw

# Deserialize from raw bytes (for stored changes)
change <- am_change_from_bytes(bytes)

# Decompose saved document into individual changes
changes <- am_load_changes(bytes) # From am_save() output

# Check document completeness
missing <- am_get_missing_deps(doc)
missing <- am_get_missing_deps(doc, heads) # With specific heads

# R → Automerge
doc <- as_automerge(list(name = "Alice", age = 30L))

# Automerge → R
r_list <- from_automerge(doc)

r_list <- as.list(doc) # S3 method for documents
text_str <- as.character(text_obj) # S3 method for text objects

doc <- am_create() |>
  am_put(AM_ROOT, "name", "Alice") |>
  am_put(AM_ROOT, "age", 30L) |>
  am_commit("Initial data")

# Save to file
writeBin(am_save(doc), "document.automerge")

# Load from file
doc <- am_load(readBin("document.automerge", "raw", 1e6))

AM_ROOT # Root object (NULL)
AM_OBJ_TYPE_LIST # "list"
AM_OBJ_TYPE_MAP # "map"
AM_OBJ_TYPE_TEXT # "text"

AM_MARK_EXPAND_NONE # "none" (mark doesn't expand at boundaries)
AM_MARK_EXPAND_BEFORE # "before" (expands when text inserted before start)
AM_MARK_EXPAND_AFTER # "after" (expands when text inserted after end)
AM_MARK_EXPAND_BOTH # "both" (expands at both boundaries)

# Function help
?am_create
?am_put
?am_sync

# Package help
?automerge
help(package = "automerge")

# Vignettes
vignette("automerge", "automerge")
vignette("crdt-concepts", "automerge")
vignette("sync-protocol", "automerge")
vignette(package = "automerge") # List all