featherweight / featherweight

Featherweight: Application Framework for NI LabVIEW™

Featherweight: Application Framework for LabVIEW™

Application Framework for LabVIEW™

featherweight's Issues

Refactor Backtrace to be an opaque DSHandle passed as integer

In the Router Recv loop in C, one DSHandle can be re-used, since it is copied during the PostLVUserEvent; it does not even need memcpy() or moveblock() into LabVIEW

FTW-Regex-Execute.vi does not protect against NULL ptr input (e.g., cannot Run the VI)

What is an Inbox for?

These are open-ended questions that might shed a lot of light on one of the fundamental assumptions FTW appears to make about Actors: Every actor comes with a predefined request-type message interface when it gets launched.

1. I understand that the Inbox is a parallelized REP socket with some extra features and logic bundled in. Can you enumerate those features and their utility in FTW?
2. Why not just create an Actor template that has a SyncServer loop for inbound messages and a separate job-worker loop? Why abstract the message processing loop into a dynamically launched VI that the framework maintains on its own?
3. Are there any known situations where a developer wants to customize an actor's Inbox but can't because it's baked into the framework?
4. Is it reasonably straightforward to create an Actor and attach a SyncServer or Subscriber to it alongside the Inbox? Is it reasonable to destroy the Inbox event registration when the actor launches and use a different messaging paradigm like Subscriber, instead?

Investigate if the Type test can load two types from vi.lib dynamically rather than statically linking

FTW-Time-Display-Timestamp.vi uses localized decimal separator for ISO-8601

On my machine I use comma as a decimal separator by default. E.g. This is a problem when inserting inserting a record that contains timestamp into SQLite database and SQLite expect timestamp in ISO-8601 format.

Set/Get Arrays in JSON object

The JSON API doesn't appear to support arrays yet. Do you have a timeline for adding that feature?

Clean up branches; switch from `expressions` to `master`

Note to all who have forked; this repository will soon be switching from the expressions branch to master as the default branch, and many development branches will be removed.

This is an effort to clean up the repository, but will temporarily break your workflow. To fix things, simply switch to the master branch in your local clone.

Comment here if you have any issues after this change is made! I'll keep the issue open for a few days after the switch.

Refactor Return Code VI to unify all of the FTW kernel with common error codes

Perhaps replace "Prototype" VIs in XNodes with static code defined in the GenerateCode abilities

Nominally, this code would exist within a Diagram Disable structure.

Benefits include:

• Tighter coupling between these two VIs, which are already extremely tightly coupled
• Less code
• Likely faster generation; copying a single GObject rather than many, and removing overhead of opening and scraping that VI ref

Add `"SupervisorOwner": true` to the Actor Instance Configuration

This enables Top-Level actors to be the "Owner" of a Supervisor, meaning that it is created prior to the actor launching, and also shuts down after the actor closes. This prevents the bookkeeping of needing to introduce it at the top-level, and the additional "weirdness" of that actor instance effectively having two connections to the Supervisor.

Investigate putting Enum in cluster types for strong typing to avoid things like wiring wrong socket types

Rename template case to "Actor-Specific Static Dependencies and Unused Terminals"

Enable SQLite Fetch to have a single scalar input to output an array of that type

This is shorthand notation for having an array of clusters where the cluster has only a single element

SQLite test fails due to string localization of decimal point (found with Andrej)

Create "Ask" XNode for the Connector library

Create "End of Subject String Reached" boolean output on Execute Regex

SQLite Interrupt from contributor AP

Create "Launch Actor" XNode

Need SQLite test case for IEEE 754 "special" values

Pretty much just like the JSON test: https://github.com/featherweight/featherweight/blob/master/tests/core/FTW-Test-JSON-FloatingPoint.vi

Related to #49

Make all error terminals like `FTW-ERR-Assert-Truthy.vi`

SQL Fetch XNode returns -42 when SQL contains bound variable not in data

This run-time error could be due to application developer error, so -42 should not be thrown.

Refactor away from LVClass for the Supervisor to avoid potential link errors when building into executable

Check into "Close if Lonely" or "Purgative" for Inbox Callback

FTW-System-CommandLineExecute.vi invokes a Windows function on any OS

This VI invokes the "SystemExec_Win" function from LabVIEW but doesn't check which OS it's running on. Calling the VI on an RT Linux target results in this error while trying to deploy code to the target:

Deploying FTW-System-CommandLineExecute.viFTW-System-CommandLineExecute.vi loaded with errors on the target and was closed.
LabVIEW: (Hex 0x627) The function name for the LabVIEW:SystemExec_Win:C node cannot be found in the library. To correct this error, right-click the Call Library Function Node and select Configure from the shortcut menu. Then choose the correct function name.

Need ability comment out jobs in scheduler sequence

Why does FTW include its own PCRE library?

What doesn't the one built into LabVIEW provide?

Need to set NN_REQ_RESEND_IVL, especially for PoisonPill (library default is 60sec)

JSON Fetch needs to accept object (instead of just string)

Need to unconditionally connect to Launcher in Handshake (currently error is wired into creation of this socket)

This will also fix issues with the Actor Stress benchmark

Savepoint instead of transaction in SQLite insert

Wrap as savepoint should be used instead of wrap as transaction in SQLite insert since it has more benefits (e.g. it can be nested).

Add Jobs and Messages to the Supervisor log as a new table

Define error codes in terms of `0xBAD00000` codes (see #23)

Need JSON read from file

VI conpanes are inconsistent

Some public VIs use the 4224 conpane, and others use 5335. One example is the "FTW-Actor-GetNextJob.vi" (4224) and "FTW-Actor-OpeningHandshake.vi" (5335), both public methods on the FTW-Actor-Instance class. Ideally, all full-sized public VIs should use one conpane or the other.

FTW-Test-JSON-Deserialize.vi can't be deployed to cRIO

I'm trying to run the test on a cRIO-9064. Before deploying code, an error is presented that says, "VI is a member of a project library type that is not supported for remote deployment and will not be downloaded."

The VI name that's cut off in this image is FTW-JSON-Fetch.xnode:FTW-JSON-Core-GetFlattenedValue.vi.

Cannot flatten Path (and perhaps other types) in VIConPaneInterface Reflection

... since the default flattened type is not all zeroes.

Create typedef for Regex callout and consider which parameters should be exposed

Rename JobSequencer to JobScheduler

Refactor Inbox away from Event Callbacks to Event Handler Structure and Semaphores

Since it turns out Event Callbacks are not threadsafe in LabVIEW :-(

Consider leaving Row Type as Recommended on DB Fetch; return 2D array of strings if unwired

Need Message Routing Rules

For instance,

  "MessageRoutingRules":
    {
      "FTW: Poison Pill":
        {
          "QueueSize": -1,
          "Ignore": false,
          "Lossy": false
        },
      "PING":
        {
          "QueueSize": -1,
          "Ignore": false,
          "AutomaticResponse": "PONG",
          "Lossy": false
        },
      "Default":
        {
          "QueueSize": -1,
          "BatchSize": 10,
          "BatchTimeout": 100,
          "Ignore": false,
          "MaxSize": 16384,
          "Action": "DropClient"
        }

SQLite returns 0.0 when it should return NaN

See: https://github.com/featherweight/ftw-kernel/blob/master/sqlite/ftw_sqlite.c#L203-L206

Socket Libraries need icons

SQLite Insert XNode should inherit control display format as means of serializing inputs

While speaking with @andrejpanjan about the SQLite Insert XNode ...

Specifically, we were on the topic of the Timestamp datatype. How should this be serialized when inserted into the database?

There exists at least three distinct use-cases:

• dates (no time)
• absolute timestamp
• relative (elapsed) time

However, the XNode must choose one. We could potentially:

• create another option on the XNode (an input that determines serialization format of the Timestamp type)
• create a pre-roll setting of that DB connection, which is another method that caches the preferred display format in the object (à la the JSON format function ... which is a related story)
• encode a special signal in the data name of the element, such as postpending the data name with [timestamp] or prepending with #date or such

But ideally, the general situation is being able to detect the display format of the data itself. This way, it would be serialized exactly as it is written on the source type definition (e.g., a constant on the diagram, or a control on the panel) (where, the use of the word 'type definition' here is distinct from Type Definition)

One potential issue here is that we'd need to do some upstream type propagation, because I don't think this display format is transmitted on the wire.

Remove Regex class

Put original regex pattern in the object so callout indices can extract callout names

Re-evaluate putting `Launch` and `Shutdown` into Supervisor Connector

See: 47ad171

Create XNode to filter to only unique items in an array

Create new class operator CodeGen callback for cleaning up broken/bendy wires

msvcr120.dll is missing

On a fresh machine, I cloned this repository and opened FTW.lvproj in LabVIEW 2015. While loading the project, LV threw this error dialog:

It appears this is a dependency of ftw32.dll on Windows. I worked around the issue by manually downloading the VS2013 C++ redistributable and installing it on my machine from here: https://www.microsoft.com/en-us/download/details.aspx?id=40784

This should be distributed with the build outputs of the ftw-kernel project and included in the featherweight repo's dependencies.