Class: Parse::Schema::IndexMigrator
- Inherits:
-
Object
- Object
- Parse::Schema::IndexMigrator
- Defined in:
- lib/parse/schema/index_migrator.rb
Overview
Reconciliation engine for ‘Parse::Core::Indexing` declarations vs. the actual MongoDB index state. Reads existing indexes via the reader connection (so `plan` works in dry-run mode without writer config); applies via the writer connection through MongoDB.create_index / MongoDB.drop_index (which re-check the triple gate and run their own per-call idempotency check on the writer-side index list).
Multi-collection. A single model can declare indexes against both its own collection (via ‘mongo_index`) and one or more `_Join:<field>:<ParentClass>` collections (via `mongo_relation_index`). `plan` returns a Hash keyed by collection name with one entry per unique target collection across the declaration list. `apply!` returns a similarly-keyed result Hash.
Parse-managed indexes (the ones Parse Server auto-creates on collections like ‘_User`, `_Session`, `_Role`) are never proposed for drop, regardless of whether they appear in the declaration list. The list is conservative — any name matching PARSE_MANAGED_INDEX_PATTERNS is treated as off-limits to the migrator, full stop.
Constant Summary collapse
- PARSE_MANAGED_INDEX_PATTERNS =
Names / patterns of indexes Parse Server creates and owns. The migrator excludes these from both ‘to_drop` (so a missing declaration never proposes their removal) and from `in_sync` (so they don’t visually clutter operator review). They appear in ‘parse_managed:` for transparency.
**Coverage is not forward-compatible.** This list reflects the indexes Parse Server auto-creates as of Parse Server 7.x. Any future Parse Server release that adds a new managed index will cause that index to be classified as an orphan and be eligible for drop under ‘DROP=true`. Operators upgrading Parse Server should re-review this list before re-running `parse:mongo:indexes:apply` with the drop flag.
Any non-declared, non-managed index — including DBA-created diagnostic indexes, indexes created by other Parse SDKs, and MongoDB Atlas index recommendations — is also classified as an orphan. If you need to preserve such an index, declare it via Core::Indexing#mongo_index on the model.
[ /\A_id_\z/, /\A_username_unique\z/, /\A_email_unique\z/, /\Aemail_1\z/, /\Ausername_1\z/, /\A_session_token_/, /\A_email_verify_token_/, /\A_perishable_token_/, /\A_account_lockout_/, /\Acase_insensitive_/, ].freeze
Instance Attribute Summary collapse
-
#model_class ⇒ Object
readonly
Returns the value of attribute model_class.
Instance Method Summary collapse
-
#apply!(drop: false) ⇒ Hash{String => Hash}
Apply the plan across all target collections.
-
#apply_for!(collection, drop: false) ⇒ Object
Per-collection apply.
-
#collection_name ⇒ String
The model’s primary collection name (parse_class).
-
#initialize(model_class) ⇒ IndexMigrator
constructor
A new instance of IndexMigrator.
-
#plan ⇒ Hash{String => Hash}
Compute the plan: what would change if ‘apply!` ran now.
-
#plan_for(collection) ⇒ Hash
Per-collection plan.
-
#target_collections ⇒ Array<String>
Unique target collections across all declarations.
Constructor Details
#initialize(model_class) ⇒ IndexMigrator
Returns a new instance of IndexMigrator.
62 63 64 65 66 67 |
# File 'lib/parse/schema/index_migrator.rb', line 62 def initialize(model_class) unless model_class.is_a?(Class) && model_class < Parse::Object raise ArgumentError, "IndexMigrator expects a Parse::Object subclass; got #{model_class.inspect}" end @model_class = model_class end |
Instance Attribute Details
#model_class ⇒ Object (readonly)
Returns the value of attribute model_class.
60 61 62 |
# File 'lib/parse/schema/index_migrator.rb', line 60 def model_class @model_class end |
Instance Method Details
#apply!(drop: false) ⇒ Hash{String => Hash}
Apply the plan across all target collections. Additive by default; ‘drop: true` opts into orphan removal on every target. Each drop carries its own confirmation envelope through `Parse::MongoDB.drop_index`.
151 152 153 154 155 |
# File 'lib/parse/schema/index_migrator.rb', line 151 def apply!(drop: false) target_collections.each_with_object({}) do |coll, h| h[coll] = apply_for!(coll, drop: drop) end end |
#apply_for!(collection, drop: false) ⇒ Object
Per-collection apply. Honors the same triple-gate / idempotency rules as the cross-collection ‘apply!`. When `drop: true` the method runs orphan drops BEFORE create_index so freed index slots are available to satisfy `to_create` — required when the collection is at or near the 64-index cap. Capacity is checked against the post-drop count, matching the actual mid-apply state.
164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 |
# File 'lib/parse/schema/index_migrator.rb', line 164 def apply_for!(collection, drop: false) p = plan_for(collection) capacity_ok = drop ? p[:capacity_ok_with_drop] : p[:capacity_ok] return { created: [], skipped_exists: [], dropped: [], conflicts: p[:conflicts], capacity_blocked: true } unless capacity_ok created = [] # Pre-seed skipped_exists with the declarations the plan already # classified as in_sync — they don't go through create_index, but # callers expect the result to reflect EVERY declaration's fate. skipped = p[:in_sync].dup dropped = [] # Drops run BEFORE creates so a full collection with one orphan # and one new declaration doesn't hit "too many indexes" before # the drop frees a slot. if drop p[:orphans].each do |name| confirm = "drop:#{collection}:#{name}" res = Parse::MongoDB.drop_index(collection, name, confirm: confirm, allow_system_classes: collection.start_with?("_Join:")) dropped << name if res == :dropped end end p[:to_create].each do |decl| result = Parse::MongoDB.create_index( collection, decl[:keys], name: decl[:options][:name], unique: decl[:options][:unique] == true, sparse: decl[:options][:sparse] == true, partial_filter: decl[:options][:partial_filter], expire_after: decl[:options][:expire_after], allow_system_classes: collection.start_with?("_Join:"), ) (result == :exists ? skipped : created) << decl end { created: created, skipped_exists: skipped, dropped: dropped, conflicts: p[:conflicts], capacity_blocked: false, } end |
#collection_name ⇒ String
Returns the model’s primary collection name (parse_class).
70 71 72 |
# File 'lib/parse/schema/index_migrator.rb', line 70 def collection_name @model_class.parse_class end |
#plan ⇒ Hash{String => Hash}
Compute the plan: what would change if ‘apply!` ran now.
86 87 88 89 90 |
# File 'lib/parse/schema/index_migrator.rb', line 86 def plan target_collections.each_with_object({}) do |coll, h| h[coll] = plan_for(coll) end end |
#plan_for(collection) ⇒ Hash
Per-collection plan. Filters declarations to those targeting ‘collection`, then runs the diff against the actual MongoDB state for that collection.
Capacity accounting reports two scenarios so callers can reason about both apply modes from a single plan:
- `:capacity_after`, `:capacity_ok` — additive-only mode
(no drops). Equal to `used + to_create.size`.
- `:capacity_after_with_drop`, `:capacity_ok_with_drop` —
additive + orphan removal. Equal to `used + to_create.size
- orphans.size`. Use these when planning an `apply!(drop:
true)` call.
107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 |
# File 'lib/parse/schema/index_migrator.rb', line 107 def plan_for(collection) existing = fetch_existing_indexes(collection) declared = declarations_for(collection) managed, ours = partition_parse_managed(existing) to_create, in_sync, conflicts = diff_declarations(declared, ours) declared_names = declared.map { |d| d[:options][:name] }.compact.to_set declared_sigs = declared.map { |d| key_sig(d[:keys]) }.to_set orphans = ours.reject do |idx| declared_sigs.include?(key_sig(idx["key"] || idx[:key])) || declared_names.include?(idx["name"] || idx[:name]) end max = Parse::Core::Indexing::MAX_INDEXES_PER_COLLECTION used = existing.size after_no_drop = used + to_create.size after_with_drop = after_no_drop - orphans.size { collection: collection, declared: declared, existing: existing, parse_managed: managed.map { |i| i["name"] || i[:name] }, to_create: to_create, in_sync: in_sync, conflicts: conflicts, orphans: orphans.map { |i| i["name"] || i[:name] }.compact, capacity_used: used, capacity_after: after_no_drop, capacity_remaining: max - after_no_drop, capacity_ok: after_no_drop <= max, capacity_after_with_drop: after_with_drop, capacity_remaining_with_drop: max - after_with_drop, capacity_ok_with_drop: after_with_drop <= max, } end |
#target_collections ⇒ Array<String>
Returns unique target collections across all declarations. Includes the parent collection only when at least one declaration targets it (i.e. a non-relation index).
77 78 79 |
# File 'lib/parse/schema/index_migrator.rb', line 77 def target_collections @model_class.mongo_index_declarations.map { |d| d[:collection] || collection_name }.uniq end |