(core) control the distribution of attachment metadata

Summary:
for users who don't automatically have deep rights
to the document, provide them with attachment metadata only
for rows they have access to. This is a little tricky to
do efficiently. We provide attachment metadata when an
individual table is fetched, rather than on initial document
load, so we don't block that load on a full document scan.
We provide attachment metadata to a client when we see that
we are shipping rows mentioning particular attachments,
without making any effort to keep track of the metadata they
already have.

Test Plan: updated tests

Reviewers: dsagal, jarek

Reviewed By: dsagal, jarek

Differential Revision: https://phab.getgrist.com/D3722
This commit is contained in:
Paul Fitzpatrick
2022-12-21 11:40:00 -05:00
parent 6dce083484
commit 472a9a186e
14 changed files with 429 additions and 169 deletions

View File

@@ -1,5 +1,5 @@
import {ActionGroup} from 'app/common/ActionGroup';
import {CellValue, TableDataAction, UserAction} from 'app/common/DocActions';
import {BulkAddRecord, CellValue, TableDataAction, UserAction} from 'app/common/DocActions';
import {FormulaProperties} from 'app/common/GranularAccessClause';
import {UIRowId} from 'app/common/UIRowId';
import {FetchUrlOptions, UploadResult} from 'app/common/uploads';
@@ -137,14 +137,30 @@ export interface QueryFilters {
// - empty: value should be falsy (e.g. null) or an empty list, filters is ignored
export type QueryOperation = "in" | "intersects" | "empty";
/**
* Results of fetching a table. Includes the table data you would
* expect. May now also include attachment metadata referred to in the table
* data. Attachment data is expressed as a BulkAddRecord, since it is
* not a complete table, just selected rows. Attachment data is
* currently included in fetches when (1) granular access control is
* in effect, and (2) the user is neither an owner nor someone with
* read access to the entire document, and (3) there is an attachment
* column in the fetched table. This is exactly what the standard
* Grist client needs, but in future it might be desirable to give
* more control over this behavior.
*/
export interface TableFetchResult {
tableData: TableDataAction;
attachments?: BulkAddRecord;
}
/**
* Response from useQuerySet(). A query returns data AND creates a subscription to receive
* DocActions that affect this data. The querySubId field identifies this subscription, and must
* be used in a disposeQuerySet() call to unsubscribe.
*/
export interface QueryResult {
export interface QueryResult extends TableFetchResult {
querySubId: number; // ID of the subscription, to use with disposeQuerySet.
tableData: TableDataAction;
}
/**
@@ -229,7 +245,7 @@ export interface ActiveDocAPI {
/**
* Fetches a particular table from the data engine to return to the client.
*/
fetchTable(tableId: string): Promise<TableDataAction>;
fetchTable(tableId: string): Promise<TableFetchResult>;
/**
* Fetches the generated Python code for this document. (TODO rename this misnomer.)

View File

@@ -0,0 +1,82 @@
import { AddRecord, BulkAddRecord, BulkRemoveRecord, BulkUpdateRecord,
getColIdsFromDocAction, getColValuesFromDocAction,
getTableId, RemoveRecord, ReplaceTableData, TableDataAction,
UpdateRecord } from 'app/common/DocActions';
import { DocData } from 'app/common/DocData';
import { isNumber } from 'app/common/gutil';
/**
* Represent current attachment columns as a map from tableId to a set of
* colIds.
*/
export type AttachmentColumns = Map<string, Set<string>>;
/**
* Enumerate attachment columns, represented as a map from tableId to
* a set of colIds.
*/
export function getAttachmentColumns(metaDocData: DocData): AttachmentColumns {
const tablesTable = metaDocData.getMetaTable('_grist_Tables');
const columnsTable = metaDocData.getMetaTable('_grist_Tables_column');
const attachmentColumns: Map<string, Set<string>> = new Map();
for (const column of columnsTable.filterRecords({type: 'Attachments'})) {
const table = tablesTable.getRecord(column.parentId);
const tableId = table?.tableId;
if (!tableId) {
/* should never happen */
throw new Error('table not found');
}
if (!attachmentColumns.has(tableId)) {
attachmentColumns.set(tableId, new Set());
}
attachmentColumns.get(tableId)!.add(column.colId);
}
return attachmentColumns;
}
/**
* Get IDs of attachments that are present in attachment columns in an action.
*/
export function gatherAttachmentIds(
attachmentColumns: AttachmentColumns,
action: AddRecord | BulkAddRecord | UpdateRecord | BulkUpdateRecord |
RemoveRecord | BulkRemoveRecord | ReplaceTableData | TableDataAction
): Set<number> {
const tableId = getTableId(action);
const attColumns = attachmentColumns.get(tableId);
const colIds = getColIdsFromDocAction(action) || [];
const attIds = new Set<number>();
if (!attColumns || !colIds.some(colId => attColumns.has(colId))) {
return attIds;
}
for (const colId of colIds) {
if (!attColumns.has(colId)) { continue; }
const values = getColValuesFromDocAction(action, colId);
if (!values) { continue; }
for (const v of values) {
// We expect an array. What should we do with other types?
// If we were confident no part of Grist would interpret non-array
// values as attachment ids, then we should let them be added, as
// part of Grist's spreadsheet-style willingness to allow invalid
// data. I decided to go ahead and require that numbers or number-like
// strings should be checked as if they were attachment ids, just in
// case. But if this proves awkward for someone, it could be reasonable
// to only check ids in an array after confirming Grist is strict in
// how it interprets material in attachment cells.
if (typeof v === 'number') {
attIds.add(v);
} else if (Array.isArray(v)) {
for (const p of v) {
if (typeof p === 'number') {
attIds.add(p);
}
}
} else if (typeof v === 'boolean' || v === null) {
// Nothing obvious to do here.
} else if (isNumber(v)) {
attIds.add(Math.round(parseFloat(v)));
}
}
}
return attIds;
}

View File

@@ -178,9 +178,12 @@ export function toTableDataAction(tableId: string, colValues: TableColValues): T
// Convert from TableDataAction (used mainly by the sandbox) to TableColValues (used by DocStorage
// and external APIs).
export function fromTableDataAction(tableData: TableDataAction): TableColValues {
const rowIds: number[] = tableData[2];
const colValues: BulkColValues = tableData[3];
// Also accepts a TableDataAction nested as a tableData member of a larger structure,
// for convenience in dealing with the result of fetches.
export function fromTableDataAction(tableData: TableDataAction|{tableData: TableDataAction}): TableColValues {
const data = ('tableData' in tableData) ? tableData.tableData : tableData;
const rowIds: number[] = data[2];
const colValues: BulkColValues = data[3];
return {id: rowIds, ...colValues};
}
@@ -203,3 +206,33 @@ export function getColValues(records: Partial<RowRecord>[]): BulkColValues {
}
return result;
}
/**
* Extract the col ids mentioned in a record-related DocAction as a list
* (even if the action is not a bulk action). Returns undefined if no col ids
* mentioned.
*/
export function getColIdsFromDocAction(docActions: RemoveRecord | BulkRemoveRecord | AddRecord |
BulkAddRecord | UpdateRecord | BulkUpdateRecord | ReplaceTableData |
TableDataAction): string[] | undefined {
if (docActions[3]) { return Object.keys(docActions[3]); }
return undefined;
}
/**
* Extract column values for a particular column as CellValue[] from a
* record-related DocAction. Undefined if absent.
*/
export function getColValuesFromDocAction(docAction: RemoveRecord | BulkRemoveRecord | AddRecord |
BulkAddRecord | UpdateRecord | BulkUpdateRecord | ReplaceTableData |
TableDataAction, colId: string): CellValue[]|undefined {
const colValues = docAction[3];
if (!colValues) { return undefined; }
const cellValues = colValues[colId];
if (!cellValues) { return undefined; }
if (Array.isArray(docAction[2])) {
return cellValues as CellValue[];
} else {
return [cellValues as CellValue];
}
}

View File

@@ -9,24 +9,39 @@ import {schema, SchemaTypes} from 'app/common/schema';
import fromPairs = require('lodash/fromPairs');
import groupBy = require('lodash/groupBy');
import {ActionDispatcher} from './ActionDispatcher';
import {TableFetchResult} from './ActiveDocAPI';
import {
BulkColValues, ColInfo, ColInfoWithId, ColValues, DocAction,
RowRecord, TableDataAction
} from './DocActions';
import {ColTypeMap, MetaRowRecord, MetaTableData, TableData} from './TableData';
type FetchTableFunc = (tableId: string) => Promise<TableDataAction>;
type FetchTableFunc = (tableId: string) => Promise<TableFetchResult>;
export class DocData extends ActionDispatcher {
private _tables: Map<string, TableData> = new Map();
private _fetchTableFunc: (tableId: string) => Promise<TableDataAction>;
/**
* If metaTableData is not supplied, then any tables needed should be loaded manually,
* using syncTable(). All column types will be set to Any, which will affect default
* values.
*/
constructor(private _fetchTableFunc: FetchTableFunc, metaTableData: {[tableId: string]: TableDataAction} | null) {
constructor(fetchTableFunc: FetchTableFunc, metaTableData: {[tableId: string]: TableDataAction} | null) {
super();
// Wrap fetchTableFunc slightly to handle any extra attachment data that
// may come along for the ride.
this._fetchTableFunc = async (tableId: string) => {
const {tableData, attachments} = await fetchTableFunc(tableId);
if (attachments) {
// Back-end doesn't keep track of which attachments we already have,
// so there may be duplicates of rows we already have - but happily
// BulkAddRecord overwrites duplicates now.
this.receiveAction(attachments);
}
return tableData;
};
if (metaTableData === null) { return; }
// Create all meta tables, and populate data we already have.
for (const tableId in schema) {

View File

@@ -2,7 +2,8 @@
* TableData maintains a single table's data.
*/
import {ActionDispatcher} from 'app/common/ActionDispatcher';
import {BulkColValues, CellValue, ColInfo, ColInfoWithId, ColValues, DocAction,
import {
BulkAddRecord, BulkColValues, CellValue, ColInfo, ColInfoWithId, ColValues, DocAction,
isSchemaAction, ReplaceTableData, RowRecord, TableDataAction} from 'app/common/DocActions';
import {getDefaultForType} from 'app/common/gristTypes';
import {arrayRemove, arraySplice, getDistinctValues} from 'app/common/gutil';
@@ -250,16 +251,40 @@ export class TableData extends ActionDispatcher implements SkippableRows {
/**
* Return data in TableDataAction form ['TableData', tableId, [...rowIds], {...}]
* Optionally takes a list of row ids to return data from. If a row id is
* not actually present in the table, a row of nulls will be returned for it.
*/
public getTableDataAction(): TableDataAction {
const rowIds = this.getRowIds();
public getTableDataAction(desiredRowIds?: number[]): TableDataAction {
const rowIds = desiredRowIds || this.getRowIds();
let bulkColValues: {[colId: string]: CellValue[]};
if (desiredRowIds) {
const len = rowIds.length;
bulkColValues = {};
for (const colId of this.getColIds()) { bulkColValues[colId] = Array(len); }
for (let i = 0; i < len; i++) {
const index = this._rowMap.get(rowIds[i]);
for (const {colId, values} of this._colArray) {
const value = (index === undefined) ? null : values[index];
bulkColValues[colId][i] = value;
}
}
} else {
bulkColValues = fromPairs(
this.getColIds()
.filter(colId => colId !== 'id')
.map(colId => [colId, this.getColValues(colId)! as CellValue[]]));
}
return ['TableData',
this.tableId,
rowIds as number[],
fromPairs(
this.getColIds()
.filter(colId => colId !== 'id')
.map(colId => [colId, this.getColValues(colId)! as CellValue[]]))];
bulkColValues];
}
public getBulkAddRecord(desiredRowIds?: number[]): BulkAddRecord {
const tableData = this.getTableDataAction(desiredRowIds?.sort((a, b) => a - b));
return [
'BulkAddRecord', tableData[1], tableData[2], tableData[3],
];
}
/**
@@ -357,6 +382,13 @@ export class TableData extends ActionDispatcher implements SkippableRows {
// ---- The following methods implement ActionDispatcher interface ----
protected onAddRecord(action: DocAction, tableId: string, rowId: number, colValues: ColValues): void {
if (this._rowMap.get(rowId) !== undefined) {
// If adding a record that already exists, act like an update.
// We rely on this behavior for distributing attachment
// metadata.
this.onUpdateRecord(action, tableId, rowId, colValues);
return;
}
const index: number = this._rowIdCol.length;
this._rowMap.set(rowId, index);
this._rowIdCol[index] = rowId;
@@ -366,14 +398,28 @@ export class TableData extends ActionDispatcher implements SkippableRows {
}
protected onBulkAddRecord(action: DocAction, tableId: string, rowIds: number[], colValues: BulkColValues): void {
const index: number = this._rowIdCol.length;
let destIndex: number = this._rowIdCol.length;
for (let i = 0; i < rowIds.length; i++) {
this._rowMap.set(rowIds[i], index + i);
this._rowIdCol[index + i] = rowIds[i];
}
for (const {colId, defl, values} of this._colArray) {
for (let i = 0; i < rowIds.length; i++) {
values[index + i] = colValues.hasOwnProperty(colId) ? colValues[colId][i] : defl;
const srcIndex = this._rowMap.get(rowIds[i]);
if (srcIndex !== undefined) {
// If adding a record that already exists, act like an update.
// We rely on this behavior for distributing attachment
// metadata.
for (const colId in colValues) {
if (colValues.hasOwnProperty(colId)) {
const colData = this._columns.get(colId);
if (colData) {
colData.values[srcIndex] = colValues[colId][i];
}
}
}
} else {
this._rowMap.set(rowIds[i], destIndex);
this._rowIdCol[destIndex] = rowIds[i];
for (const {colId, defl, values} of this._colArray) {
values[destIndex] = colValues.hasOwnProperty(colId) ? colValues[colId][i] : defl;
}
destIndex++;
}
}
}