Client-side editing: if code changed whilst offline, upload and compile once connection to server returns

[zdateh]

Current Client-Side behaviour when "online" (connected to IRIS server):

1. Edit routine/class
2. Save
3. Change is uploaded to server and compiled

Current Client-Side behaviour when "offline" (not connected to IRIS server):

1. Edit routine/class
2. Save
3. Error message during upload as no connection to server

Makes sense, but once connection to the server is reestablished, your options to re-trigger the failed upload-to-server-and-compile seem to be:

1. Save any previously changed routines/classes again (have to remember what you've changed offline - prone to failure)
2. Do an 'Import and Compile' of your entire workspace (takes ages as it's uploading and recompiling EVERYTHING, when you only want to upload the changes)
3. Switch branches away, and then back to your working branch - this seems to trigger an upload and compile of everything unique to the branch (works if using Git, but a bit of a hack)

Unless I've missed a better solution, I would like to suggest that once the connection to the sever is re-established, any changes saved offline are uploaded and compiled. Thanks :)

[isc-bsaviano]

@zdateh Can you share what error message you're seeing? If the server connection has "active": false then VS Code should not attempt to the sync the file on save. Do you make the server "offline" another way?

This is an interesting idea, but I think it would be hard to implement. We would need to persist "dirty" files in a way that would survive VS Code restarts. You'd also need some way to protect against changing server connections entirely. What if a user never makes the server connection active? Do we keep these records forever?

[zdateh]

@zdateh Can you share what error message you're seeing? If the server connection has "active": false then VS Code should not attempt to the sync the file on save. Do you make the server "offline" another way?

the server is "offline" because my local machine can't reach it - eg, required vpn is off or wifi is off. our vpn goes down fairly frequently!

This is an interesting idea, but I think it would be hard to implement. We would need to persist "dirty" files in a way that would survive VS Code restarts.

maybe check what files in the workspace have timestamps later than the last time files were sent to the server?

You'd also need some way to protect against changing server connections entirely. What if a user never makes the server connection active? Do we keep these records forever?

following the above suggestion, perhaps clear last_time_files_sent (or whatever) if server reconnected to is different to that when connection was lost.

right now losing server connection at any point halts the ability of the local machine to be the 'golden source', which at some point I can predict is going to cause us to get into a mess, unless we make it some kind of common practice to do an 'import all'.

[isc-bsaviano]

Your local copy is still the "golden source" because it is tracked by source control. If you're having VPN issues then I recommend developing against a local instance instead. The Community edition makes that easy to set up. When editing local files I like to think of IRIS as an SDK, like the JDK or Node.js. You always want your SDK to be available when you're developing.

[zdateh]

I agree but I'm 20 years deep working for various intersystems development houses and it's always been local studio hooking into remote database :) I could personally setup CE but I doubt I'll get buy-in from the wider team, during our transition to vscode.

I suppose I could revert back to server-side development via vscode, but client-side has so many advantages especially around source control.

[isc-bsaviano]

Server-side editing would be much worse experience for you. You couldn't do anything if the server connection went down. I understand the reasons for this request, and we made some major improvements in this area for version 3. I'll leave this open until I can find time to discuss it with the other maintainers.

[jer-war]

I know best practice these days is for a local Docker to be the server with IRIS running. But a lot of people still use IRIS on Windows for the database and this is therefore often on a separate server from their local machine.
When using VS code with Git locally in this scenario, that's when you can end up with having a difference between the local code and the code in IRIS (VPN disconnects/branch switching/switching namespaces/etc.).
It would be great if there was a way to do some sort of check or comparison of my current Git branch to what's in the IRIS namespace.
Or a way to deploy/import all, perhaps by creating an XML and pushing via the API to IRIS to load and compile.
Or a way for the local IRIS Lite terminal to be able to read from local file system so you could Import and CompileAll for the local directory.

[isc-bsaviano]

I know best practice these days is for a local Docker to be the server with IRIS running. But a lot of people still use IRIS on Windows for the database and this is therefore often on a separate server from their local machine.

I understand the history and why people work the way that they do. I also understand that switching can be hard. However, since the Community Edition and containers have come along it's never been easier to have a private instance for local development.

When using VS code with Git locally in this scenario, that's when you can end up with having a difference between the local code and the code in IRIS

This shouldn't be a major problem though. The "source of truth" for source control purposes is your local folder, which will have all the changes. The server is only used for compilation, intellisense, and running unit tests. You can still edit without the server connection and none of your work would be lost.

It would be great if there was a way to do some sort of check or comparison of my current Git branch to what's in the IRIS namespace.
Or a way to deploy/import all, perhaps by creating an XML and pushing via the API to IRIS to load and compile

You can do this. Just right-click on the folder in the Files Explorer and select Import and Compile. I have already reduced the number of times you'd need to do this for version 3 (I watch the file system for changes that happen outside of VS Code, like Git branch changes). The request in this issue is for the manual import to not be necessary if the server goes down while you're editing.

Or a way for the local IRIS Lite terminal to be able to read from local file system so you could Import and CompileAll for the local directory.

It's not possible for a remote server to read your local file system, unless that server is a container running on your local machine and you've mounted the directory.

[jer-war]

Thanks a lot for your response. The Import and Compile (for the whole directory) seems to work perfectly compared to a few months ago, so that's really useful for these scenarios.