templates
Statement templates: a statement built once per cache key and reused with new values on every later call.
Most of the cost of a simple read is building the statement: ormar walks the
filters, joins, selected fields and ordering, and SQLAlchemy assembles the
expression tree. Two calls that differ only in their values build the same
tree, so QuerySet builds it once with named bind parameters ("slots") in
place of the values and keeps it in a cache of the model::
| Text Only | |
|---|---|
1 2 3 4 5 6 | |
The cache key (see QuerySet.select_cache_key) holds everything that
changes the SQL text and none of the bound values: filtered columns,
operators, and/or/not nesting, select_related, selected fields, ordering
and whether a limit or offset is present. Values SQLAlchemy writes into the
SQL text instead of binding (None as IS NULL, booleans, isnull)
are part of the key as well.
Writes are cached the same way. Model.save(), update(), delete(),
load() and upsert() run one statement per model and set of written
columns, binding the values as ormar_v{i} (in sorted column order) and the
primary key as ormar_pk::
| Text Only | |
|---|---|
1 2 3 | |
QuerySet.update() and delete() bind both filter and value slots, and
bulk_create and bulk_update reuse one statement per set of columns for
all their rows.
Running the same statement object every time also lets the native compiler reuse its plan for it instead of compiling the statement again.
The package is split by role:
cache- the per-model template cachecache_keys- what makes two querysets share a statementslots- named bind parameters and the values collected per callreads- statements wrapped around a select, clean-up of templateswrites- insert and upsert templates (updates, deletes and selects by primary key:ormar.queryset.write_statements)
SlotBinder
Turns per-call values of a read statement into named bind parameters while
the statement is built. A filter action's value is named ormar_p{n}
after its position among the canonical actions, limit and offset get fixed
names::
| Text Only | |
|---|---|
1 2 3 | |
Slots are looked up by action object, not by order of use, because one
action can be emitted twice: with select_related and a limit, the
filters go into the pagination subquery and the outer query, and both
copies bind the same ormar_p{n}, so one value fills both.
Passed down through Query, FilterQuery, LimitQuery and
OffsetQuery; without a binder they bind the plain values as before.
Source code in ormar/queryset/templates/slots.py
| Python | |
|---|---|
162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 | |
__init__(actions)
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
actions
|
List[FilterAction]
|
filter actions of the queryset in canonical order |
required |
Source code in ormar/queryset/templates/slots.py
| Python | |
|---|---|
182 183 184 185 186 187 188 189 | |
bind(action, value)
Return the named bind parameter for a filter action's value.
The parameter is built the way SQLAlchemy builds the bind of
column <op> value: it is compared to the column's type, so the value
gets the column's bind processing (dates, JSON, enums and so on), and
an in filter is expanding, so it renders one placeholder per item of
the list it is given on each call.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
action
|
FilterAction
|
filter action the value belongs to |
required |
value
|
Any
|
value to bind |
required |
Returns:
| Type | Description |
|---|---|
BindParameter
|
named bind parameter |
Source code in ormar/queryset/templates/slots.py
| Python | |
|---|---|
191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 | |
bind_key_in(action, value, column_types)
Return the expanding bind parameter of a row-value IN over several
key columns, typed as a tuple of the column types so each part gets its
column's bind processing::
| Text Only | |
|---|---|
1 2 | |
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
action
|
FilterAction
|
the key |
required |
value
|
Any
|
key values, one tuple of parts per key |
required |
column_types
|
List[Any]
|
types of the key columns, in key order |
required |
Returns:
| Type | Description |
|---|---|
BindParameter
|
named expanding bind parameter |
Source code in ormar/queryset/templates/slots.py
| Python | |
|---|---|
217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 | |
limit(value)
Return the named bind parameter for a limit, typed as an integer like
the one SQLAlchemy creates for select.limit(10).
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
value
|
int
|
limit value |
required |
Returns:
| Type | Description |
|---|---|
BindParameter
|
named bind parameter |
Source code in ormar/queryset/templates/slots.py
| Python | |
|---|---|
244 245 246 247 248 249 250 251 252 253 254 255 256 | |
offset(value)
Return the named bind parameter for an offset, typed as an integer like
the one SQLAlchemy creates for select.offset(10).
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
value
|
int
|
offset value |
required |
Returns:
| Type | Description |
|---|---|
BindParameter
|
named bind parameter |
Source code in ormar/queryset/templates/slots.py
| Python | |
|---|---|
258 259 260 261 262 263 264 265 266 267 268 269 270 | |
StatementTemplate
Bases: NamedTuple
A statement from the template cache and the values to run it with for one call::
| Text Only | |
|---|---|
1 2 3 | |
Source code in ormar/queryset/templates/cache.py
| Python | |
|---|---|
31 32 33 34 35 36 37 38 39 40 41 42 | |
TemplateCache
Least-recently-used cache of one model's statements, keyed by cache key.
The size is bounded so an application that builds queries from many
different filter combinations cannot grow memory without limit: once
max_size keys are cached, adding one drops the key used longest ago.
Source code in ormar/queryset/templates/cache.py
| Python | |
|---|---|
45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 | |
__init__(max_size=TEMPLATE_CACHE_SIZE)
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
max_size
|
int
|
number of cache keys kept |
TEMPLATE_CACHE_SIZE
|
Source code in ormar/queryset/templates/cache.py
| Python | |
|---|---|
54 55 56 57 58 59 60 | |
clear()
Remove every cached statement.
Source code in ormar/queryset/templates/cache.py
| Python | |
|---|---|
98 99 100 | |
get(key)
Return the statement cached for a key, or None, and mark it as the most recently used.
Another thread using the same model may evict the entry between reading it and marking it, so that case is ignored: the statement is still valid for this call.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
key
|
Any
|
cache key of the statement |
required |
Returns:
| Type | Description |
|---|---|
Any
|
cached statement or None |
Source code in ormar/queryset/templates/cache.py
| Python | |
|---|---|
62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 | |
put(key, statement)
Cache the statement of a key. When the cache is full, the key used longest ago is dropped.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
key
|
Any
|
cache key of the statement |
required |
statement
|
Any
|
statement to cache |
required |
Source code in ormar/queryset/templates/cache.py
| Python | |
|---|---|
84 85 86 87 88 89 90 91 92 93 94 95 96 | |
action_cache_key(action)
Return everything about a filter action that changes the SQL text: the
field path, the table alias of the joined relation, the operator, the
kind of value (see value_cache_key) and whether LIKE needs an ESCAPE
clause for % or _ in the value.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
action
|
FilterAction
|
filter action |
required |
Returns:
| Type | Description |
|---|---|
Tuple[Any, ...]
|
hashable cache key of the action |
Source code in ormar/queryset/templates/cache_keys.py
| Python | |
|---|---|
60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 | |
aggregate_statement(select, func_name, select_columns)
Wrap a read select into aggregate function columns, the statement run by
QuerySet.max(), min(), sum() and avg()::
| Text Only | |
|---|---|
1 2 | |
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
select
|
Select
|
read select of the queryset |
required |
func_name
|
str
|
aggregate function name |
required |
select_columns
|
List[Any]
|
aggregate column expressions |
required |
Returns:
| Type | Description |
|---|---|
Executable
|
aggregate statement |
Source code in ormar/queryset/templates/reads.py
| Python | |
|---|---|
69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 | |
build_insert(table, names)
Build an insert of the given columns with named value binds::
| Text Only | |
|---|---|
1 | |
The names are sorted by the caller, so the same columns given in any order share one statement.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
table
|
Table
|
written table |
required |
names
|
Tuple[str, ...]
|
written column names, sorted |
required |
Returns:
| Type | Description |
|---|---|
Executable
|
insert statement |
Source code in ormar/queryset/templates/writes.py
| Python | |
|---|---|
45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 | |
build_update_assignment(model, on_update, update_values)
Build the MySQL SET of the user variables a multi-row upsert reads its
on_update values and the bound update_values from (see
update_assignment).
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
model
|
Type[Any]
|
written model |
required |
on_update
|
Tuple[str, ...]
|
columns updated with their |
required |
update_values
|
Mapping[str, Any]
|
value or expression node by column name |
required |
Returns:
| Type | Description |
|---|---|
Executable
|
assignment statement |
Source code in ormar/queryset/templates/writes.py
| Python | |
|---|---|
147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 | |
cached_statement(model_cls, key, builder, **builder_kwargs)
Return the statement cached for a write cache key, building it on a miss.
The statement is marked as a template (TEMPLATE_OPTION), so the native
compiler keeps its plan even when SQLAlchemy cannot cache it.
A write's cache key is small (the kind of statement and the written column
names), so callers pass the key directly. The statement is only built on
a miss, from builder(**builder_kwargs)::
| Text Only | |
|---|---|
1 2 3 4 5 6 7 | |
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
model_cls
|
Type[Any]
|
model the statement writes |
required |
key
|
Tuple[Any, ...]
|
cache key of the write |
required |
builder
|
Callable[..., Executable]
|
builds the statement from |
required |
builder_kwargs
|
Any
|
keyword arguments of the builder |
{}
|
Returns:
| Type | Description |
|---|---|
Executable
|
statement for the key |
Source code in ormar/queryset/templates/cache.py
| Python | |
|---|---|
127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 | |
canonical_actions(filter_clauses, exclude_clauses)
Return every filter action of a queryset in canonical order: filter clauses first, then exclude clauses, each in build order::
| Text Only | |
|---|---|
1 2 | |
SlotBinder (naming the slots while building) and template_params
(collecting values on every call) both use this list, so a value always
ends up in the slot built for it.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
filter_clauses
|
List[Any]
|
queryset filter clauses |
required |
exclude_clauses
|
List[Any]
|
queryset exclude clauses |
required |
Returns:
| Type | Description |
|---|---|
List[FilterAction]
|
filter actions in canonical order |
Source code in ormar/queryset/templates/slots.py
| Python | |
|---|---|
135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 | |
clauses_cache_key(clauses)
Return the cache key of a list of filter actions and filter groups, keeping
the and/or/not nesting, since or_(a, b) and and_(a, b) build
different SQL from the same actions::
| Text Only | |
|---|---|
1 2 | |
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
clauses
|
Iterable[Any]
|
filter actions and filter groups |
required |
Returns:
| Type | Description |
|---|---|
Tuple[Any, ...]
|
hashable cache key of the clauses |
Source code in ormar/queryset/templates/cache_keys.py
| Python | |
|---|---|
81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 | |
count_statement(select, distinct, pk_column_names)
Wrap a read select into a row count, the statement run by
QuerySet.count(). With distinct the rows are grouped by primary key
first, so a model joined to several related rows is counted once::
| Text Only | |
|---|---|
1 2 3 | |
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
select
|
Select
|
read select of the queryset |
required |
distinct
|
bool
|
count distinct primary keys only |
required |
pk_column_names
|
Tuple[str, ...]
|
primary key column aliases of the model |
required |
Returns:
| Type | Description |
|---|---|
Executable
|
count statement |
Source code in ormar/queryset/templates/reads.py
| Python | |
|---|---|
38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 | |
excludable_cache_key(excludable)
Return the fields included and excluded per model and relation, which
decide the selected columns (fields(), exclude_fields(),
values()). Field sets are sorted so the same selection always gives
the same key.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
excludable
|
ExcludableItems
|
queryset's fields to include and exclude |
required |
Returns:
| Type | Description |
|---|---|
Tuple[Any, ...]
|
hashable cache key of the projection |
Source code in ormar/queryset/templates/cache_keys.py
| Python | |
|---|---|
109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 | |
exists_statement(select)
Wrap a read select into SELECT EXISTS (...), the statement run by
QuerySet.exists().
This and the other *_statement helpers are the wrap callables of
QuerySet.build_select_template: they run on the select once, when a
cache key is first seen, and the wrapped statement is what gets cached.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
select
|
Select
|
read select of the queryset |
required |
Returns:
| Type | Description |
|---|---|
Executable
|
exists statement |
Source code in ormar/queryset/templates/reads.py
| Python | |
|---|---|
21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 | |
insert_statement(model_cls, names, on_conflict=None, conflict_target=None, update_fields=None, on_update=(), returning=(), update_values=None, update_variables=False)
Return the insert of a model for the given sorted column names, from the
model's template cache. save() binds one row to it and
bulk_create runs it with executemany for every row with the same
columns.
With on_conflict the insert skips or updates conflicting rows (see
build_conflict_insert): the target defaults to the primary key and the
updated columns to the inserted ones but the target, the key and
on_update, or to none with update_values. One statement is cached
per shape of update_values (see update_values_key).
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
model_cls
|
Type[Any]
|
inserted model |
required |
names
|
Tuple[str, ...]
|
inserted column names, sorted |
required |
on_conflict
|
Optional[str]
|
|
None
|
conflict_target
|
Optional[Sequence[str]]
|
names of the fields of the conflicting unique key |
None
|
update_fields
|
Optional[Sequence[str]]
|
names of the fields updated from the inserted row |
None
|
on_update
|
Tuple[str, ...]
|
columns updated with their |
()
|
returning
|
Tuple[str, ...]
|
columns of the written row to return, if any |
()
|
update_values
|
Optional[Mapping[str, Any]]
|
value or expression node set by column name |
None
|
update_variables
|
bool
|
read the bound update values from MySQL user
variables set by |
False
|
Returns:
| Type | Description |
|---|---|
Executable
|
insert statement with |
Source code in ormar/queryset/templates/writes.py
| Python | |
|---|---|
251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 | |
insert_template(model_cls, values, **statement_kwargs)
Return the insert of a model for the given column values, used by
save() and one-row bulk_create::
| Text Only | |
|---|---|
1 2 3 | |
Rows that leave out the primary key or a column with a server default write a different set of columns, so each set gets its own statement.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
model_cls
|
Type[Any]
|
inserted model |
required |
values
|
Mapping[str, Any]
|
values by column name |
required |
statement_kwargs
|
Any
|
conflict handling (see |
{}
|
Returns:
| Type | Description |
|---|---|
StatementTemplate
|
statement and values to bind |
Source code in ormar/queryset/templates/writes.py
| Python | |
|---|---|
362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 | |
iter_actions(clauses)
Yield the filter actions of a clause list in the order their SQL is built:
a group's nested groups first, then its own actions (the order
FilterGroup.get_text_clause emits them in)::
| Text Only | |
|---|---|
1 2 | |
The order is the same for every queryset of one cache key, so an action's index here can name its slot.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
clauses
|
List[Union[FilterAction, FilterGroup]]
|
filter actions and filter groups |
required |
Returns:
| Type | Description |
|---|---|
Iterator[FilterAction]
|
filter actions in build order |
Source code in ormar/queryset/templates/slots.py
| Python | |
|---|---|
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 | |
order_cache_key(order_bys)
Return the cache key of the requested orderings: field path, direction, table alias and nulls placement of each, in order::
| Text Only | |
|---|---|
1 | |
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
order_bys
|
Iterable[OrderAction]
|
order actions of the query |
required |
Returns:
| Type | Description |
|---|---|
Tuple[Any, ...]
|
hashable cache key of the ordering |
Source code in ormar/queryset/templates/cache_keys.py
| Python | |
|---|---|
133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 | |
release_values(statement)
Drop the values of a template's named slots, so a cached statement does not keep the first call's values (possibly large lists or documents) alive for as long as it stays cached. Every call binds its own values anyway.
SQLAlchemy may have cloned the bind parameters while wrapping the select
(for example into a subquery), so the whole statement is walked instead of
clearing only the parameters SlotBinder created.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
statement
|
Executable
|
built template statement |
required |
Source code in ormar/queryset/templates/reads.py
| Python | |
|---|---|
92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 | |
template_cache_for(model_cls)
Return the statement template cache of a model class, creating it on the
model's first read. setdefault keeps a single cache when two threads
create it at the same time.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
model_cls
|
Type[Any]
|
model class |
required |
Returns:
| Type | Description |
|---|---|
TemplateCache
|
the model's template cache |
Source code in ormar/queryset/templates/cache.py
| Python | |
|---|---|
110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 | |
template_params(actions, limit, offset)
Collect the values a template binds for one call, keyed like the slots
SlotBinder names::
| Text Only | |
|---|---|
1 2 | |
Actions whose value is written into the SQL text (see
FilterAction.binds_slot) get no entry, but still use up their index,
so the names of the other slots do not depend on those values. The offset
is only bound when non-zero, matching OffsetQuery, which leaves out
OFFSET 0.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
actions
|
List[FilterAction]
|
filter actions in canonical order |
required |
limit
|
Optional[int]
|
effective limit (applied when not None) |
required |
offset
|
Optional[int]
|
effective offset (applied when truthy) |
required |
Returns:
| Type | Description |
|---|---|
Dict[str, Any]
|
values per slot name |
Source code in ormar/queryset/templates/slots.py
| Python | |
|---|---|
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 | |
update_values_key(update_values)
Return the shape of the update_values of an upsert for cache keys: the
column of a value, the column and expression_key of an expression, so
the literals do not split the cache.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
update_values
|
Mapping[str, Any]
|
value or expression node by column name |
required |
Returns:
| Type | Description |
|---|---|
Tuple[Any, ...]
|
hashable shape |
Source code in ormar/queryset/templates/writes.py
| Python | |
|---|---|
128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 | |
update_values_params(update_values)
Return the values bound by the update_values of an upsert, named like
update_value_clauses names them::
| Text Only | |
|---|---|
1 2 | |
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
update_values
|
Mapping[str, Any]
|
value or expression node by column name |
required |
Returns:
| Type | Description |
|---|---|
Dict[str, Any]
|
values by bind parameter name |
Source code in ormar/queryset/templates/writes.py
| Python | |
|---|---|
107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 | |
value_binds(table, names)
Return named bind parameters for the written columns, ormar_v{i} in
the given column order, typed like their columns::
| Text Only | |
|---|---|
1 2 | |
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
table
|
Table
|
written table |
required |
names
|
Tuple[str, ...]
|
written column names, sorted |
required |
Returns:
| Type | Description |
|---|---|
Dict[Any, Any]
|
bind parameter per column |
Source code in ormar/queryset/templates/slots.py
| Python | |
|---|---|
273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 | |
value_cache_key(action)
Return the part of a filter value that changes the SQL text::
| Text Only | |
|---|---|
1 2 3 4 5 6 | |
Values written into the SQL text are part of the key as they are. For a bound value only its type is kept, so values of different types never share a statement.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
action
|
FilterAction
|
filter action |
required |
Returns:
| Type | Description |
|---|---|
Any
|
hashable value kind |
Source code in ormar/queryset/templates/cache_keys.py
| Python | |
|---|---|
28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 | |
value_params(names, values)
Return the written values keyed by their bind parameter names, matching
value_binds for the same names::
| Text Only | |
|---|---|
1 2 | |
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
names
|
Tuple[str, ...]
|
written column names, sorted |
required |
values
|
Mapping[str, Any]
|
written values by column name |
required |
Returns:
| Type | Description |
|---|---|
Dict[str, Any]
|
values by bind parameter name |
Source code in ormar/queryset/templates/slots.py
| Python | |
|---|---|
296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 | |