mirror of
https://github.com/gristlabs/grist-core.git
synced 2024-10-27 20:44:07 +00:00
a56714e1ab
Summary: Trigger formulas can be calculated for new records, or for new records and updates to certain fields, or all fields. They do not recalculate on open, and they MAY be set directly by the user, including for data-cleaning. - Column metadata now includes recalcWhen and recalcDeps fields. - Trigger formulas are NOT recalculated on open or on schema changes. - When recalcWhen is "never", formula isn't calculated even for new records. - When recalcWhen is "allupdates", formula is calculated for new records and any manual (non-formula) updates to the record. - When recalcWhen is "", formula is calculated for new records, and changes to recalcDeps fields (which may be formula fields or column itself). - A column whose recalcDeps includes itself is a "data-cleaning" column; a value set by the user will still trigger the formula. - All trigger-formulas receive a "value" argument (to support the case above). Small changes - Update RefLists (used for recalcDeps) when target rows are deleted. - Add RecordList.__contains__ (for `rec in refList` or `id in refList` checks) - Clarify that Calculate action has replaced load_done() in practice, and use it in tests too, to better match reality. Left for later: - UI for setting recalcWhen / recalcDeps. - Implementation of actions such as "Recalculate for all cells". - Allowing trigger-formulas access to the current user's info. Test Plan: Added a comprehensive python-side test for various trigger combinations Reviewers: paulfitz, alexmojaki Reviewed By: paulfitz Differential Revision: https://phab.getgrist.com/D2872
270 lines
11 KiB
Python
270 lines
11 KiB
Python
import six
|
|
|
|
import actions
|
|
import logger
|
|
import schema
|
|
from objtypes import strict_equal
|
|
|
|
log = logger.Logger(__name__, logger.INFO)
|
|
|
|
class DocActions(object):
|
|
def __init__(self, engine):
|
|
self._engine = engine
|
|
|
|
#----------------------------------------
|
|
# Actions on records.
|
|
#----------------------------------------
|
|
|
|
def AddRecord(self, table_id, row_id, column_values):
|
|
self.BulkAddRecord(
|
|
table_id, [row_id], {key: [val] for key, val in six.iteritems(column_values)})
|
|
|
|
def BulkAddRecord(self, table_id, row_ids, column_values):
|
|
table = self._engine.tables[table_id]
|
|
for row_id in row_ids:
|
|
assert row_id not in table.row_ids, \
|
|
"docactions.[Bulk]AddRecord for existing record #%s" % row_id
|
|
|
|
self._engine.out_actions.undo.append(actions.BulkRemoveRecord(table_id, row_ids).simplify())
|
|
self._engine.out_actions.summary.add_records(table_id, row_ids)
|
|
|
|
self._engine.add_records(table_id, row_ids, column_values)
|
|
|
|
def RemoveRecord(self, table_id, row_id):
|
|
return self.BulkRemoveRecord(table_id, [row_id])
|
|
|
|
def BulkRemoveRecord(self, table_id, row_ids):
|
|
table = self._engine.tables[table_id]
|
|
|
|
# Ignore records that don't exist in the table.
|
|
row_ids = [r for r in row_ids if r in table.row_ids]
|
|
if not row_ids:
|
|
return
|
|
|
|
# Collect the undo values, and unset all values in the column (i.e. set to defaults), just to
|
|
# make sure we don't have stale values hanging around.
|
|
undo_values = {}
|
|
for column in six.itervalues(table.all_columns):
|
|
if not column.is_private() and column.col_id != "id":
|
|
col_values = [column.raw_get(r) for r in row_ids]
|
|
default = column.getdefault()
|
|
# If this column had all default values, don't include it into the undo BulkAddRecord.
|
|
if not all(strict_equal(val, default) for val in col_values):
|
|
undo_values[column.col_id] = col_values
|
|
for row_id in row_ids:
|
|
column.unset(row_id)
|
|
|
|
# Generate the undo action.
|
|
self._engine.out_actions.undo.append(
|
|
actions.BulkAddRecord(table_id, row_ids, undo_values).simplify())
|
|
self._engine.out_actions.summary.remove_records(table_id, row_ids)
|
|
|
|
# Invalidate the deleted rows, so that anything that depends on them gets recomputed.
|
|
self._engine.invalidate_records(table_id, row_ids)
|
|
|
|
def UpdateRecord(self, table_id, row_id, columns):
|
|
self.BulkUpdateRecord(
|
|
table_id, [row_id], {key: [val] for key, val in six.iteritems(columns)})
|
|
|
|
def BulkUpdateRecord(self, table_id, row_ids, columns):
|
|
table = self._engine.tables[table_id]
|
|
for row_id in row_ids:
|
|
assert row_id in table.row_ids, \
|
|
"docactions.[Bulk]UpdateRecord for non-existent record #%s" % row_id
|
|
|
|
# Load the updated values.
|
|
undo_values = {}
|
|
for col_id, values in six.iteritems(columns):
|
|
col = table.get_column(col_id)
|
|
undo_values[col_id] = [col.raw_get(r) for r in row_ids]
|
|
for (row_id, value) in zip(row_ids, values):
|
|
col.set(row_id, value)
|
|
|
|
# Non-formula columns may get invalidated and recalculated if they have a trigger formula.
|
|
# Prevent such recalculation if we set an explicit value for them (we want to prevent it
|
|
# even if triggered by something else within the same useraction).
|
|
if not col.is_formula():
|
|
self._engine.prevent_recalc(col.node, row_ids, should_prevent=True)
|
|
|
|
# Generate the undo action.
|
|
self._engine.out_actions.undo.append(
|
|
actions.BulkUpdateRecord(table_id, row_ids, undo_values).simplify())
|
|
|
|
# Invalidate the updated rows, just for the columns that got changed (and, as always,
|
|
# anything that depends on them).
|
|
self._engine.invalidate_records(table_id, row_ids, col_ids=columns.keys())
|
|
|
|
|
|
def ReplaceTableData(self, table_id, row_ids, column_values):
|
|
old_data = self._engine.fetch_table(table_id, formulas=False)
|
|
self._engine.out_actions.undo.append(actions.ReplaceTableData(*old_data))
|
|
self._engine.out_actions.summary.remove_records(table_id, old_data[1])
|
|
self._engine.out_actions.summary.add_records(table_id, row_ids)
|
|
self._engine.load_table(actions.TableData(table_id, row_ids, column_values))
|
|
|
|
#----------------------------------------
|
|
# Actions on columns.
|
|
#----------------------------------------
|
|
|
|
def AddColumn(self, table_id, col_id, col_info):
|
|
table = self._engine.tables[table_id]
|
|
assert not table.has_column(col_id), "Column %s already exists in %s" % (col_id, table_id)
|
|
|
|
# Add the new column to the schema object maintained in the engine.
|
|
self._engine.schema[table_id].columns[col_id] = schema.dict_to_col(col_info, col_id=col_id)
|
|
self._engine.rebuild_usercode()
|
|
self._engine.new_column_name(table)
|
|
|
|
# Generate the undo action.
|
|
self._engine.out_actions.undo.append(actions.RemoveColumn(table_id, col_id))
|
|
self._engine.out_actions.summary.add_column(table_id, col_id)
|
|
|
|
def RemoveColumn(self, table_id, col_id):
|
|
table = self._engine.tables[table_id]
|
|
assert table.has_column(col_id), "Column %s not in table %s" % (col_id, table_id)
|
|
|
|
# Generate (if needed) the undo action to restore the data.
|
|
undo_action = None
|
|
column = table.get_column(col_id)
|
|
if not column.is_private():
|
|
default = column.getdefault()
|
|
# Add to undo a BulkUpdateRecord for non-default values in the column being removed.
|
|
undo_values = [(r, column.raw_get(r)) for r in table.row_ids
|
|
if not strict_equal(column.raw_get(r), default)]
|
|
|
|
# Remove the specified column from the schema object.
|
|
colinfo = self._engine.schema[table_id].columns.pop(col_id)
|
|
self._engine.rebuild_usercode()
|
|
|
|
# Generate the undo action(s); if for a formula column, add them to the calc summary.
|
|
if undo_values:
|
|
if column.is_formula():
|
|
changes = [(r, v, default) for (r, v) in undo_values]
|
|
self._engine.out_actions.summary.add_changes(table_id, col_id, changes)
|
|
else:
|
|
row_ids = [r for (r, v) in undo_values]
|
|
values = [v for (r, v) in undo_values]
|
|
undo_action = actions.BulkUpdateRecord(table_id, row_ids, {col_id: values}).simplify()
|
|
self._engine.out_actions.undo.append(undo_action)
|
|
|
|
self._engine.out_actions.undo.append(actions.AddColumn(
|
|
table_id, col_id, schema.col_to_dict(colinfo, include_id=False)))
|
|
self._engine.out_actions.summary.remove_column(table_id, col_id)
|
|
|
|
def RenameColumn(self, table_id, old_col_id, new_col_id):
|
|
table = self._engine.tables[table_id]
|
|
|
|
assert table.has_column(old_col_id), "Column %s not in table %s" % (old_col_id, table_id)
|
|
assert not table.has_column(new_col_id), \
|
|
"Column %s already exists in %s" % (new_col_id, table_id)
|
|
old_column = table.get_column(old_col_id)
|
|
|
|
# Replace the renamed column in the schema object.
|
|
schema_table_info = self._engine.schema[table_id]
|
|
colinfo = schema_table_info.columns.pop(old_col_id)
|
|
schema_table_info.columns[new_col_id] = schema.SchemaColumn(
|
|
new_col_id, colinfo.type, colinfo.isFormula, colinfo.formula)
|
|
|
|
self._engine.rebuild_usercode()
|
|
self._engine.new_column_name(table)
|
|
|
|
# We replaced the old column with a new Column object (not strictly necessary, but simpler).
|
|
# For a raw data column, we need to copy over the data from the old column object.
|
|
new_column = table.get_column(new_col_id)
|
|
new_column.copy_from_column(old_column)
|
|
|
|
# Generate the undo action.
|
|
self._engine.out_actions.undo.append(actions.RenameColumn(table_id, new_col_id, old_col_id))
|
|
self._engine.out_actions.summary.rename_column(table_id, old_col_id, new_col_id)
|
|
|
|
def ModifyColumn(self, table_id, col_id, col_info):
|
|
table = self._engine.tables[table_id]
|
|
assert table.has_column(col_id), "Column %s not in table %s" % (col_id, table_id)
|
|
old_column = table.get_column(col_id)
|
|
|
|
# Modify the specified column in the schema object.
|
|
schema_table_info = self._engine.schema[table_id]
|
|
old = schema_table_info.columns[col_id]
|
|
new = schema.SchemaColumn(col_id,
|
|
col_info.get('type', old.type),
|
|
bool(col_info.get('isFormula', old.isFormula)),
|
|
col_info.get('formula', old.formula))
|
|
if new == old:
|
|
log.info("ModifyColumn called which was a noop")
|
|
return
|
|
|
|
undo_col_info = {k: v for k, v in six.iteritems(schema.col_to_dict(old, include_id=False))
|
|
if k in col_info}
|
|
|
|
# Remove the column from the schema, then re-add it, to force creation of a new column object.
|
|
schema_table_info.columns.pop(col_id)
|
|
self._engine.rebuild_usercode()
|
|
|
|
schema_table_info.columns[col_id] = new
|
|
self._engine.rebuild_usercode()
|
|
|
|
# Fill in the new column with the values from the old column.
|
|
new_column = table.get_column(col_id)
|
|
for row_id in table.row_ids:
|
|
new_column.set(row_id, old_column.raw_get(row_id))
|
|
|
|
# Generate the undo action.
|
|
self._engine.out_actions.undo.append(actions.ModifyColumn(table_id, col_id, undo_col_info))
|
|
|
|
#----------------------------------------
|
|
# Actions on tables.
|
|
#----------------------------------------
|
|
|
|
def AddTable(self, table_id, columns):
|
|
assert table_id not in self._engine.tables, "Table %s already exists" % table_id
|
|
|
|
# Update schema, and re-generate the module code.
|
|
self._engine.schema[table_id] = schema.SchemaTable(table_id, schema.dict_list_to_cols(columns))
|
|
self._engine.rebuild_usercode()
|
|
|
|
# Generate the undo action.
|
|
self._engine.out_actions.undo.append(actions.RemoveTable(table_id))
|
|
self._engine.out_actions.summary.add_table(table_id)
|
|
|
|
def RemoveTable(self, table_id):
|
|
assert table_id in self._engine.tables, "Table %s doesn't exist" % table_id
|
|
|
|
# Create undo actions to restore all the data records of this table.
|
|
table_data = self._engine.fetch_table(table_id, formulas=True)
|
|
undo_action = actions.BulkAddRecord(*table_data).simplify()
|
|
if undo_action:
|
|
self._engine.out_actions.undo.append(undo_action)
|
|
|
|
# Update schema, and re-generate the module code.
|
|
schema_table = self._engine.schema.pop(table_id)
|
|
self._engine.rebuild_usercode()
|
|
|
|
# Generate the undo action.
|
|
self._engine.out_actions.undo.append(actions.AddTable(
|
|
table_id, schema.cols_to_dict_list(schema_table.columns)))
|
|
self._engine.out_actions.summary.remove_table(table_id)
|
|
|
|
def RenameTable(self, old_table_id, new_table_id):
|
|
assert old_table_id in self._engine.tables, "Table %s doesn't exist" % old_table_id
|
|
assert new_table_id not in self._engine.tables, "Table %s already exists" % new_table_id
|
|
|
|
old_table = self._engine.tables[old_table_id]
|
|
|
|
# Update schema, and re-generate the module code.
|
|
old = self._engine.schema.pop(old_table_id)
|
|
self._engine.schema[new_table_id] = schema.SchemaTable(new_table_id, old.columns)
|
|
self._engine.rebuild_usercode()
|
|
|
|
# Copy over all columns from the old table to the new.
|
|
new_table = self._engine.tables[new_table_id]
|
|
for new_column in six.itervalues(new_table.all_columns):
|
|
if not new_column.is_private():
|
|
new_column.copy_from_column(old_table.get_column(new_column.col_id))
|
|
new_table.grow_to_max() # We need to bring formula columns to the right size too.
|
|
|
|
# Generate the undo action.
|
|
self._engine.out_actions.undo.append(actions.RenameTable(new_table_id, old_table_id))
|
|
self._engine.out_actions.summary.rename_table(old_table_id, new_table_id)
|
|
|
|
# end
|