Summary:
This changes how user attributes are loaded. They are now loaded
directly from sqlite, with per-session caching. Optimizations
considered but not addressed yet are (1) adding indexes to user attribute
tables and (2) swapping in a thinner sqlite wrapper.
The main benefit of this diff is that changes to user attribute
tables now work. Clients whose user attributes are not changed
see no effect; clients whose user attributes have changed have
their document reloaded.
For the purposes of testing, the diff includes a tweak to
GristWSConnection to be "sticky" to a specific user when reloading
(and support machinery on the server side to honor that). Until
now, if a GristWSConnection reloads, it uses whatever the current
default user is in the cookie-based session, which can change.
This was complicating a test where multiple users were accessing
the same document via different clients with occasional document
reloads.
Code for updating when schema or rule changes happen is moved
around but not improved in any meaningful way in this diff.
Test Plan: existing tests pass; extended test
Reviewers: dsagal
Reviewed By: dsagal
Differential Revision: https://phab.getgrist.com/D2685
Summary:
This improves support for access control on document modifications. It adds:
* Checking of create/remove/update access for row-level changes.
* Use of `newRec` variable in formulas.
It is now possible to have distinct clients with read+write access to different rows of the same table.
This is another incremental step. There are deficiencies in actions that include schema changes, and many other lacunae. But the overall flow is taking shape.
Access control is done at the DocAction level, requiring the sandbox to process the UserActions, and then be reverted if the action proves unlawful. This could be optimized away in many simple and important cases, but I'm not sure it is possible to avoid in general.
Test Plan: added tests
Reviewers: dsagal
Reviewed By: dsagal
Differential Revision: https://phab.getgrist.com/D2677
Summary:
- Replace unicode strings with byte strings when decoding values in sandbox.
- Columns that rely on float values should derive from NumericColumn, so
that set() ensures that a float is stored even if loading an int.
- Parse unmarshallable values (['U']) into an object that can be encoded
back to the same value (rather than info a RaisedException).
- Compare NaN's as equal for deciding whether a change is a no-op.
Unrelated:
- Removed a tiny bit of unhelpful logging
Test Plan:
Added a test case that reproduces several causes of Calculate
discrepancies by loading various values into various types of formula columns.
Reviewers: paulfitz
Reviewed By: paulfitz
Differential Revision: https://phab.getgrist.com/D2676
Summary:
This implements row-level access control for outgoing messages, replacing the document reloading placeholder that was there before.
* Prior to broadcasting messages, GranularAccess is notified of actions+undo.
* While broadcasting messages to different sessions, if we find we need row level access control information, rows before and after the change are reconstructed.
* Messages are rewritten if rows that were previously forbidden are now allowed, and vice versa.
The diff is somewhat under-tested and under-optimized. Next step would be to implement row-level access control for incoming actions, which may result in some rejiggering of the code from this diff to avoid duplication of effort under some conditions.
Test Plan: added test
Reviewers: dsagal
Reviewed By: dsagal
Differential Revision: https://phab.getgrist.com/D2670
Summary:
- Added fields to _grist_ACLRules for the new Granular ACL representation
- Include a corresponding migration.
- Added ACLPermissions module with merging PermissionSets and converting to/from string.
- Implemented parsing of ACL formulas and compiling them into JS functions.
- Add automatic parsing of ACL formulas when ACLRules are added or updated.
- Convert GranularAccess to load and interpret new-style rules.
- Convert ACL UI to load and save new-style rules.
For now, no attempt to do anything better on the server or UI side, only to
reproduce previous behavior.
Test Plan: Added unittests for new files; fixed those for existing files.
Reviewers: paulfitz
Reviewed By: paulfitz
Differential Revision: https://phab.getgrist.com/D2664
Summary:
The new plans for granular access control are different and handled by
node.js. Some of the same tables will be reused, of which we never made
real use before except for expecting certain specific initial records.
This diff removes the old logic, replacing it with a stub that satisfies
the interface expected by other code.
It also removes several unused UserActions: AddUser/RemoveUser/
AddInstance/RemoveInstance.
Test Plan: Existing tests should pass.
Reviewers: paulfitz
Reviewed By: paulfitz
Differential Revision: https://phab.getgrist.com/D2662
Summary:
Loading all user data to run a migration is risky (creates more than usual
memory pressure), and almost never needed (only one migration requires it).
This diff attempts to run migrations using only metadata (_grist_* tables),
but retries if the sandbox tells it that all data is needed.
The intent is for new migrations to avoid needing all data.
Test Plan: Added a somewhat contrived unittest.
Reviewers: paulfitz
Reviewed By: paulfitz
Differential Revision: https://phab.getgrist.com/D2659
Summary:
Type conversions and formula tranforms wait for the user and bundle multiple
actions. When an unrelated action is done (e.g. adding a page widget or a
column), we want to finalize the transform before applying it.
The approach turns out fairly complicated. There is an implicit queue of
bundles (which we don't let grow beyond 2, as that's too abnormal). Bundles may
be finalized by a user clicking something, or by an unrelated action/bundle, or
(as before) by transform DOM getting disposed.
- Updated RecordLayout to use bundleActions() helper
- Added support for nesting bundleActions inside another bundle (needed for
setting visibleCol during type change)
- In an unrelated tweak, when in debug-log in ActiveDoc, use a short representation of result.
Test Plan: Added a unittest for action bundling during type transform
Reviewers: paulfitz
Reviewed By: paulfitz
Differential Revision: https://phab.getgrist.com/D2655
Summary:
- Introduce a new SQLiteDB migration, which adds DB columns for formula columns
- Newly added columns have the special ['P'] (pending) value in them
(in order to show the usual "Loading..." on the first load that triggers the migration)
- Calculated values are added to .stored/.undo fields of user actions.
- Various changes made in the sandbox to include .stored/.undo in the right order.
- OnDemand tables ignore stored formula columns, replacing them with special SQL as before
- In particular, converting to OnDemand table leaves stale values in those
columns, we should maybe clean those out.
Some tweaks on the side:
- Allow overriding chai assertion truncateThreshold with CHAI_TRUNCATE_THRESHOLD
- Rebuild python automatically in watch mode
Test Plan: Fixed various tests, updated some fixtures. Many python tests that check actions needed adjustments because actions moved from .stored to .undo. Some checks added to catch situations previously only caught in browser tests.
Reviewers: paulfitz
Reviewed By: paulfitz
Differential Revision: https://phab.getgrist.com/D2645
Summary: Adds a granular access clause for columns. Permissions can be specified for a set of columns within a table. Permissions accumulate over clauses, in a way that is intended as a placeholder pending final design.
Test Plan: Added tests. Tested manually that updates to private columns are not sent to people who don't have access to them. There are a lot of extra tests needed and TODOs to be paid down after this experimental phase.
Reviewers: dsagal
Reviewed By: dsagal
Differential Revision: https://phab.getgrist.com/D2651
Summary:
Deliberate changes:
* save snapshots to s3 prior to migrations.
* label migration snapshots in s3 metadata.
* avoid pruning migration snapshots for a month.
Opportunistic changes:
* Associate document timezone with snapshots, so pruning can respect timezones.
* Associate actionHash/Num with snapshots.
* Record time of last change in snapshots (rather than just s3 upload time, which could be a while later).
This ended up being a biggish change, because there was nowhere ideal to put tags (list of possibilities in diff).
Test Plan: added tests
Reviewers: dsagal
Reviewed By: dsagal
Differential Revision: https://phab.getgrist.com/D2646
Summary:
This is a prototype for expanding the conditions that can be used in granular ACLs.
When processing ACLs, the following variables (called "characteristics") are now available in conditions:
* UserID
* Email
* Name
* Access (owners, editors, viewers)
The set of variables can be expanded by adding a "characteristic" clause. This is a clause which specifies:
* A tableId
* The name of an existing characteristic
* A colId
The effect of the clause is to expand the available characteristics with all the columns in the table, with values taken from the record where there is a match between the specified characteristic and the specified column.
Existing clauses are generalized somewhat to demonstrate and test the use these variables. That isn't the main point of this diff though, and I propose to leave generalizing+systematizing those clauses for a future diff.
Issues I'm not dealing with here:
* How clauses combine. (The scope on GranularAccessRowClause is a hack to save me worrying about that yet).
* The full set of matching methods we'll allow.
* Refreshing row access in clients when the tables mentioned in characteristic tables change.
* Full CRUD permission control.
* Default rules (part of combination).
* Reporting errors in access rules.
That said, with this diff it is possible to e.g. assign a City to editors by their email address or name, and have only rows for those Cities be visible in their client. Ability to modify those rows, and remain updates about them, remains under incomplete control.
Test Plan: added tests
Reviewers: dsagal
Reviewed By: dsagal
Differential Revision: https://phab.getgrist.com/D2642
Summary:
This implements a form of row-level access control where for a
given table, you may specify that only owners have access to
rows for which a given column has falsy values.
For simplicity:
* Only owners may edit that table.
* Non-owners with the document open will have forced
reloads whenever the table is modified.
Baby steps...
Test Plan: added tests
Reviewers: dsagal
Reviewed By: dsagal
Differential Revision: https://phab.getgrist.com/D2633
Summary:
Add a unittest that start the server with limited memory, and build just enough
of ActionHistory to crash the server before this fix, and not after.
Test Plan: Tested manually with various memory prints, and added a test.
Reviewers: paulfitz
Reviewed By: paulfitz
Differential Revision: https://phab.getgrist.com/D2616
Summary:
* Extends `/api/docs/docId1/compare/docId2` endpoint with a `detail=1` option to include details of what changed in the document content.
* Adds an `/api/docs/docId/compare?left=HASH&right=HASH` endpoint for comparing two versions of a single document. This is needed to implement the extension to `/api/docs/docId1/compare/docId2`.
* Adds a `HashUtil` class to allow hash aliases like `HEAD` and `HEAD~`.
Everything is a bit crude:
* Changes are expressed as ActionSummary objects, which aren't fully fleshed out.
* Extra data about formula columns is inserted in an inflexible way.
This is extracted and cleaned up from https://phab.getgrist.com/D2600.
Test Plan: added tests
Reviewers: dsagal
Reviewed By: dsagal
Differential Revision: https://phab.getgrist.com/D2614
Summary:
This is an incremental step in granular access control. Using
a temporary `{colIds: '~o structure'}` representation in the
`_grist_ACLResources` table, the document structure can be set
to be controlled by owners only.
Test Plan: added test
Reviewers: dsagal
Reviewed By: dsagal
Differential Revision: https://phab.getgrist.com/D2613
Summary:
This makes it possible to serve a table or tables only to owners.
* The _grist_ACLResources table is abused (temporarily) such that rows of the form `{colId: '~o', tableId}` are interpreted as meaning that `tableId` is private to owners.
* Many websocket and api endpoints are updated to preserve the privacy of these tables.
* In a document where some tables are private, a lot of capabilities are turned off for non-owners to avoid leaking info indirectly.
* The client is tweaked minimally, to show '-' where a page with some private material would otherwise go.
No attempt is made to protect data from private tables pulled into non-private tables via formulas.
There are some known leaks remaining:
* Changes to the schema of private tables are still broadcast to all clients (fixable).
* Non-owner may be able to access snapshots or make forks or use other corners of API (fixable).
* Changing name of table makes it public, since tableId in ACLResource is not updated (fixable).
Security will require some work, the attack surface is large.
Test Plan: added tests
Reviewers: dsagal
Reviewed By: dsagal
Differential Revision: https://phab.getgrist.com/D2604
Summary: This makes the user's role (owner/editor/viewer) available in ActiveDoc methods. No use of that information is made yet, other than to log it. The bulk of the diff is getting a handle on the various ways the methods can be called, and systematizing it a bit more. In passing, access control is added to broadcasts of document changes, so users who no longer have access to a document do not receive changes if they still have the document open.
Test Plan: existing tests pass; test for broadcast access control added
Reviewers: dsagal
Reviewed By: dsagal
Differential Revision: https://phab.getgrist.com/D2599
Summary: This moves enough server material into core to run a home server. The data engine is not yet incorporated (though in manual testing it works when ported).
Test Plan: existing tests pass
Reviewers: dsagal
Reviewed By: dsagal
Differential Revision: https://phab.getgrist.com/D2552
