gristlabs_grist-core/sandbox/grist/column.py
Alex Hall 5d671bf0b3 (core) New type conversion in the backend
Summary: This is https://phab.getgrist.com/D3205 plus some changes (https://github.com/dsagal/grist/compare/type-convert...type-convert-server?expand=1) that move the conversion process to the backend. A new user action ConvertFromColumn uses `call_external` so that the data engine can delegate back to ActiveDoc. Code for creating formatters and parsers is significantly refactored so that most of the logic is in `common` and can be used in different ways.

Test Plan: The original diff adds plenty of tests.

Reviewers: georgegevoian

Reviewed By: georgegevoian

Subscribers: dsagal

Differential Revision: https://phab.getgrist.com/D3240
2022-02-04 20:28:13 +02:00

590 lines
21 KiB
Python

import json
import types
from collections import namedtuple
from numbers import Number
import six
import depend
import objtypes
import usertypes
import relabeling
import relation
import moment
import logger
from sortedcontainers import SortedListWithKey
log = logger.Logger(__name__, logger.INFO)
MANUAL_SORT = 'manualSort'
MANUAL_SORT_COL_INFO = {
'id': MANUAL_SORT,
'type': 'ManualSortPos',
'formula': '',
'isFormula': False
}
MANUAL_SORT_DEFAULT = 2147483647.0
SPECIAL_COL_IDS = {'id', MANUAL_SORT}
def is_user_column(col_id):
"""
Returns whether the col_id is of a user column (as opposed to special columns that can't be used
for user data).
"""
return col_id not in SPECIAL_COL_IDS and not col_id.startswith('#')
def is_visible_column(col_id):
"""
Returns whether this is an id of a column that's intended to be shown to the user.
"""
return is_user_column(col_id) and not col_id.startswith('gristHelper_')
def is_virtual_column(col_id):
"""
Returns whether col_id is of a special column that does not get communicated outside of the
sandbox. Lookup maps are an example.
"""
return col_id.startswith('#')
def is_validation_column_name(name):
return name.startswith("validation___")
ColInfo = namedtuple('ColInfo', ('type_obj', 'is_formula', 'method'))
def get_col_info(col_model, default_func=None):
if isinstance(col_model, types.FunctionType):
type_obj = getattr(col_model, 'grist_type', usertypes.Any())
return ColInfo(type_obj, is_formula=True, method=col_model)
else:
return ColInfo(col_model, is_formula=False, method=col_model.default_func or default_func)
class BaseColumn(object):
"""
BaseColumn holds a column of data, whether raw or computed.
"""
def __init__(self, table, col_id, col_info):
self.type_obj = col_info.type_obj
self._data = []
self.col_id = col_id
self.table_id = table.table_id
self.node = depend.Node(self.table_id, col_id)
self._is_formula = col_info.is_formula
self._is_private = bool(col_info.method) and getattr(col_info.method, 'is_private', False)
self.update_method(col_info.method)
# Always initialize to include the special empty record at index 0.
self.growto(1)
def update_method(self, method):
"""
After rebuilding user code, we reuse existing column objects, but need to replace their
'method' function. The method may refer to variables in the generated "usercode" module, and
it's important that all such references are to the rebuilt "usercode" module.
"""
self.method = method
def is_formula(self):
"""
Whether this is a formula column. Note that a non-formula column may have an associated
method, which is used to fill in defaults when a record is added.
"""
return self._is_formula
def is_private(self):
"""
Returns whether this method is private to the sandbox. If so, changes to this column do not
get communicated to outside the sandbox via actions.
"""
return self._is_private
def has_formula(self):
"""
has_formula is true if formula is set, whether or not this is a formula column.
"""
return self.method is not None
def clear(self):
self._data = []
self.growto(1) # Always include the special empty record at index 0.
def destroy(self):
"""
Called when the column is deleted.
"""
del self._data[:]
def growto(self, size):
if len(self._data) < size:
self._data.extend([self.getdefault()] * (size - len(self._data)))
def size(self):
return len(self._data)
def set(self, row_id, value):
"""
Sets the value of this column for the given row_id. Value should be as returned by convert(),
i.e. of the right type, or alttext, or error (but should NOT be random wrong types).
"""
try:
self._data[row_id] = value
except IndexError:
self.growto(row_id + 1)
self._data[row_id] = value
def unset(self, row_id):
"""
Sets the value for the given row_id to the default value.
"""
self.set(row_id, self.getdefault())
def get_cell_value(self, row_id, restore=False):
"""
Returns the "rich" value for the given row_id, i.e. the value that would be seen by formulas.
E.g. for ReferenceColumns it'll be the referred-to Record object. For cells containing
alttext, this will be an AltText object. For RaisedException objects that represent a thrown
error, this will re-raise that error.
"""
raw = self.raw_get(row_id)
if isinstance(raw, objtypes.RaisedException):
# For trigger formulas, we want to restore the previous value to recalculate
# the cell one more time.
if restore and raw.has_user_input():
raw = raw.user_input
elif isinstance(raw.error, depend.CircularRefError):
# Wrapping a CircularRefError in a CellError is often redundant, but
# TODO a CellError is still useful if the calling cell is not involved in the cycle
raise raw.error
else:
raise objtypes.CellError(self.table_id, self.col_id, row_id, raw.error)
return self._convert_raw_value(raw)
def _convert_raw_value(self, raw):
if self.type_obj.is_right_type(raw):
return self._make_rich_value(raw)
return self._alt_text(raw)
def _alt_text(self, raw):
return usertypes.AltText(str(raw), self.type_obj.typename())
def _make_rich_value(self, typed_value):
"""
Called by get_cell_value() with a value of the right type for this column. Should be
implemented by derived classes to produce a "rich" version of the value.
"""
# pylint: disable=no-self-use
return typed_value
def raw_get(self, row_id):
"""
Returns the value stored for the given row_id. This may be an error or alttext, and it does
not convert to a richer object.
"""
try:
return self._data[row_id]
except IndexError:
return self.getdefault()
def safe_get(self, row_id):
"""
Returns a value of the right type, or the default value if the stored value had a wrong type.
"""
raw = self.raw_get(row_id)
return raw if self.type_obj.is_right_type(raw) else self.getdefault()
def getdefault(self):
"""
Returns the default value for this column. This is a static default; the implementation of
"default formula" logic is separate.
"""
return self.type_obj.default
def sample_value(self):
"""
Returns a sample value for this column, used for auto-completions. E.g. for a date, this
returns an actual datetime object rather than None (only its attributes should matter).
"""
return self.type_obj.default
def copy_from_column(self, other_column):
"""
Replace this column's data entirely with data from another column of the same exact type.
"""
self._data[:] = other_column._data
def convert(self, value_to_convert):
"""
Converts a value of any type to this column's type, returning either the converted value (for
which is_right_type is true), or an alttext string, or an error object.
"""
return self.type_obj.convert(value_to_convert)
def prepare_new_values(self, values, ignore_data=False, action_summary=None):
"""
This allows us to modify values and also produce adjustments to existing records. This
currently is only used by PositionColumn. Returns two lists: new_values, and
[(row_id, new_value)] list of adjustments to existing records.
If ignore_data is True, makes adjustments without regard to the existing data; this is used
for processing ReplaceTableData actions.
"""
# pylint: disable=no-self-use, unused-argument
return values, []
class DataColumn(BaseColumn):
"""
DataColumn describes a column of raw data, and holds it.
"""
pass
class ChoiceColumn(DataColumn):
def rename_choices(self, renames):
row_ids = []
values = []
for row_id, value in enumerate(self._data):
if value is not None and self.type_obj.is_right_type(value):
value = self._rename_cell_choice(renames, value)
if value is not None:
row_ids.append(row_id)
values.append(value)
return row_ids, values
def _rename_cell_choice(self, renames, value):
return renames.get(value)
class BoolColumn(BaseColumn):
def set(self, row_id, value):
# When 1 or 1.0 is loaded, we should see it as True, and similarly 0 as False. This is similar
# to how, after loading a number into a DateColumn, we should see a date, except we adjust
# booleans at set() time.
bool_value = True if value == 1 else (False if value == 0 else value)
super(BoolColumn, self).set(row_id, bool_value)
class NumericColumn(BaseColumn):
def set(self, row_id, value):
# Make sure any integers are treated as floats to avoid truncation.
# Uses `type(value) == int` rather than `isintance(value, int)` to specifically target
# ints and not bools (which are singleton instances the class int in python). But
# perhaps something should be done about bools also?
# pylint: disable=unidiomatic-typecheck
super(NumericColumn, self).set(row_id, float(value) if type(value) == int else value)
_sample_date = moment.ts_to_date(0)
_sample_datetime = moment.ts_to_dt(0, None, moment.TZ_UTC)
class DateColumn(NumericColumn):
"""
DateColumn contains numerical timestamps represented as seconds since epoch, in type float,
to midnight of specific UTC dates. Accessing them yields date objects.
"""
def _make_rich_value(self, typed_value):
return typed_value and moment.ts_to_date(typed_value)
def sample_value(self):
return _sample_date
class DateTimeColumn(NumericColumn):
"""
DateTimeColumn contains numerical timestamps represented as seconds since epoch, in type float,
and a timestamp associated with the column. Accessing them yields datetime objects.
"""
def __init__(self, table, col_id, col_info):
super(DateTimeColumn, self).__init__(table, col_id, col_info)
self._timezone = col_info.type_obj.timezone
def _make_rich_value(self, typed_value):
return typed_value and moment.ts_to_dt(typed_value, self._timezone)
def sample_value(self):
return _sample_datetime
class SafeSortKey(object):
"""
Sort key that deals with errors raised by normal comparison,
in particular for values of different types or values that can't
be compared at all (e.g. None).
This somewhat mimics the way that Python 2 compares values.
This is only needed in Python 3.
"""
__slots__ = ("value",)
def __init__(self, value):
self.value = value
def __repr__(self):
return "MixedTypesKey({self.value!r})".format(self=self)
def __eq__(self, other):
return self.value == other.value
def __lt__(self, other):
try:
return self.value < other.value
except TypeError:
if type(self.value) is type(other.value):
return id(self.value) < id(other.value)
else:
return self.type_position() < other.type_position()
def type_position(self):
# Fallback order similar to Python 2:
# - None is less than everything else
# - Numbers are less than other types
# - Other types are ordered by type name
# The first two elements use the fact that False < True (because 0 < 1)
return (
self.value is not None,
not isinstance(self.value, Number),
type(self.value).__name__,
)
if six.PY2:
_doc = SafeSortKey.__doc__
def SafeSortKey(x):
return x
SafeSortKey.__doc__ =_doc
class PositionColumn(NumericColumn):
def __init__(self, table, col_id, col_info):
super(PositionColumn, self).__init__(table, col_id, col_info)
# This is a list of row_ids, ordered by the position.
self._sorted_rows = SortedListWithKey(key=lambda x: SafeSortKey(self.raw_get(x)))
def set(self, row_id, value):
self._sorted_rows.discard(row_id)
super(PositionColumn, self).set(row_id, value)
if value != self.getdefault():
self._sorted_rows.add(row_id)
def copy_from_column(self, other_column):
super(PositionColumn, self).copy_from_column(other_column)
self._sorted_rows = SortedListWithKey(other_column._sorted_rows[:],
key=lambda x: SafeSortKey(self.raw_get(x)))
def prepare_new_values(self, values, ignore_data=False, action_summary=None):
# This does the work of adjusting positions and relabeling existing rows with new position
# (without changing sort order) to make space for the new positions. Note that this is also
# used for updating a position for an existing row: we'll find a new value for it; later when
# this value is set, the old position will be removed and the new one added.
if ignore_data:
rows = []
else:
rows = self._sorted_rows
# prepare_inserts expects floats as keys, not MixedTypesKeys
rows = SortedListWithKey(rows, key=self.raw_get)
adjustments, new_values = relabeling.prepare_inserts(rows, values)
return new_values, [(self._sorted_rows[i], pos) for (i, pos) in adjustments]
class ChoiceListColumn(ChoiceColumn):
"""
ChoiceListColumn's default value is None, but is presented to formulas as the empty list.
"""
def set(self, row_id, value):
# When a JSON string is loaded, set it to a tuple parsed from it. When a list is loaded,
# convert to a tuple to keep values immutable.
if isinstance(value, six.string_types) and value.startswith(u'['):
try:
value = tuple(json.loads(value))
except Exception:
pass
elif isinstance(value, list):
value = tuple(value)
super(ChoiceListColumn, self).set(row_id, value)
def _make_rich_value(self, typed_value):
return () if typed_value is None else typed_value
def _rename_cell_choice(self, renames, value):
if any((v in renames) for v in value):
return tuple(renames.get(choice, choice) for choice in value)
class BaseReferenceColumn(BaseColumn):
"""
Base class for ReferenceColumn and ReferenceListColumn.
"""
def __init__(self, table, col_id, col_info):
super(BaseReferenceColumn, self).__init__(table, col_id, col_info)
# We can assume that all tables have been instantiated, but not all initialized.
target_table_id = self.type_obj.table_id
self._target_table = table._engine.tables.get(target_table_id, None)
self._relation = relation.ReferenceRelation(table.table_id, target_table_id, col_id)
# Note that we need to remove these back-references when the column is removed.
if self._target_table:
self._target_table._back_references.add(self)
def destroy(self):
# Destroy the column and remove the back-reference we created in the constructor.
super(BaseReferenceColumn, self).destroy()
if self._target_table:
self._target_table._back_references.remove(self)
def _update_references(self, row_id, old_value, new_value):
raise NotImplementedError()
def set(self, row_id, value):
old = self.safe_get(row_id)
super(BaseReferenceColumn, self).set(row_id, value)
new = self.safe_get(row_id)
self._update_references(row_id, old, new)
def copy_from_column(self, other_column):
super(BaseReferenceColumn, self).copy_from_column(other_column)
self._relation.clear()
# This is hacky: we should have an interface to iterate through values of a column. (As it is,
# self._data may include values for non-existent rows; it works here because those values are
# falsy, which makes them ignored by self._update_references).
for row_id, value in enumerate(self._data):
if self.type_obj.is_right_type(value):
self._update_references(row_id, None, value)
def sample_value(self):
return self._target_table.sample_record
def get_updates_for_removed_target_rows(self, target_row_ids):
"""
Returns a list of pairs of (row_id, new_value) for values in this column that need to be
updated when target_row_ids are removed from the referenced table.
"""
affected_rows = sorted(self._relation.get_affected_rows(target_row_ids))
return [(row_id, self._raw_get_without(row_id, target_row_ids)) for row_id in affected_rows]
def _raw_get_without(self, _row_id, _target_row_ids):
"""
Returns a Ref or RefList cell value with the specified target_row_ids removed, assuming one of
them is actually present in the value. For References, it just leaves the default value.
"""
return self.getdefault()
def _lookup(self, reference_lookup, value):
col_id = (
reference_lookup.options.get("column")
or self._target_table._engine.docmodel
.get_column_rec(self.table_id, self.col_id).visibleCol.colId
or "id"
)
column = self._target_table.get_column(col_id)
# `value` is an object encoded for transmission from JS to Python,
# which is decoded to `decoded_value`.
# `raw_value` is the kind of value that would be stored in `column`.
# `rich_value` is the type of value used in formulas, especially with `lookupRecords`.
# For example, for a Date column, `raw_value` is a numerical timestamp
# and `rich_value` is a `datetime.date` object,
# assuming `value` isn't of an invalid type.
# However `value` could either be just a number
# (in which case `decoded_value` would be a number as well)
# or an encoded date (or even datetime) object like ['d', number]
# (in which case `decoded_value` would be a `datetime.date` object,
# which would get converted back to a number and then back to a date object again!)
decoded_value = objtypes.decode_object(value)
raw_value = column.convert(decoded_value)
rich_value = column._convert_raw_value(raw_value)
return self._target_table.lookup_one_record(**{col_id: rich_value})
class ReferenceColumn(BaseReferenceColumn):
"""
ReferenceColumn contains IDs of rows in another table. Accessing them yields the records in the
other table.
"""
def _make_rich_value(self, typed_value):
# If we refer to an invalid table, return integers rather than fail completely.
if not self._target_table:
return typed_value
# For a Reference, values must either refer to an existing record, or be 0. In all tables,
# the 0 index will contain the all-defaults record.
return self._target_table.Record(typed_value, self._relation)
def _update_references(self, row_id, old_value, new_value):
if old_value:
self._relation.remove_reference(row_id, old_value)
if new_value:
self._relation.add_reference(row_id, new_value)
def prepare_new_values(self, values, ignore_data=False, action_summary=None):
if action_summary and values:
values = action_summary.translate_new_row_ids(self._target_table.table_id, values)
return values, []
def convert(self, val):
if isinstance(val, objtypes.ReferenceLookup):
val = self._lookup(val, val.value) or self._alt_text(val.alt_text)
return super(ReferenceColumn, self).convert(val)
class ReferenceListColumn(BaseReferenceColumn):
"""
ReferenceListColumn maintains for each row a list of references (row IDs) into another table.
Accessing them yields RecordSets.
"""
def set(self, row_id, value):
if isinstance(value, six.string_types) and value.startswith(u'['):
try:
value = json.loads(value)
except Exception:
pass
super(ReferenceListColumn, self).set(row_id, value)
def _update_references(self, row_id, old_list, new_list):
for old_value in old_list or ():
self._relation.remove_reference(row_id, old_value)
for new_value in new_list or ():
self._relation.add_reference(row_id, new_value)
def _make_rich_value(self, typed_value):
if typed_value is None:
typed_value = []
# If we refer to an invalid table, return integers rather than fail completely.
if not self._target_table:
return typed_value
return self._target_table.RecordSet(typed_value, self._relation)
def _raw_get_without(self, row_id, target_row_ids):
"""
Returns the RefList cell value at row_id with the specified target_row_ids removed.
"""
raw = self.raw_get(row_id)
if self.type_obj.is_right_type(raw):
raw = [r for r in raw if r not in target_row_ids] or None
return raw
def convert(self, val):
if isinstance(val, objtypes.ReferenceLookup):
result = []
values = val.value
if not isinstance(values, list):
values = [values]
for value in values:
lookup_value = self._lookup(val, value)
if not lookup_value:
return self._alt_text(val.alt_text)
result.append(lookup_value)
val = result
return super(ReferenceListColumn, self).convert(val)
# Set up the relationship between usertypes objects and column objects.
usertypes.BaseColumnType.ColType = DataColumn
usertypes.Reference.ColType = ReferenceColumn
usertypes.ReferenceList.ColType = ReferenceListColumn
usertypes.Choice.ColType = ChoiceColumn
usertypes.ChoiceList.ColType = ChoiceListColumn
usertypes.DateTime.ColType = DateTimeColumn
usertypes.Date.ColType = DateColumn
usertypes.PositionNumber.ColType = PositionColumn
usertypes.Bool.ColType = BoolColumn
usertypes.Numeric.ColType = NumericColumn
def create_column(table, col_id, col_info):
return col_info.type_obj.ColType(table, col_id, col_info)