v2.4 Planning & Objectives

[jeremy-visionaid]

The counterpart ticket to v3 planning... I think we've got a decent amount of performance and correctness fixes in for v2.4. I was wondering what others were thinking about additional tickets for the v2.4 milestone?

I just have one last niggle that's on my mind, and that's to validate stream truncation. i.e. check that sectors are correctly marked as free in the FAT when the length of a stream is reduced (I'm not sure if the expectation is that the length of file should reduce in-line with the last allocated sector?). It looks like it's probably OK, so there might be nothing further to do than just convince myself that it is. Even if the file doesn't shrink automatically, then shrinking by the client seems a reasonable option anyway.

Maybe I could run through with some other analyzers too, and check that there aren't any further warnings that are worth addressing.

Other than that is v2.4.0 pretty much ready to go? 🚀

[Numpsy]

It'd be useful for me to complete #190, but the questions about that are mainly how how some error handling / validation should work and how 'clever' it should be (simple works for my usage).

[Numpsy]

Maybe I could run through with some other analyzers too, and check that there aren't any further warnings that are worth addressing.

There's a set of suggestions from NetAnalyzers to seal a set of the TypedPropertyValue subclasses - e.g.

I don't know how much real benefit it'd be, but they are all private types so it's not a breaking API change to do it

[ironfede]

I think we are almost ready for 2.4.0...
At this point I would only target correctness and solidity of the package.

[Numpsy]

I've done a quick run with benchmarkdotnet against my own project just reading document properties from a Word document and got this:

2.3 openmcdf release

| Method                  | Job      | Runtime  | Mean      | Error    | StdDev   | Gen0     | Gen1    | Allocated  |
|------------------------ |--------- |--------- |----------:|---------:|---------:|---------:|--------:|-----------:|
| CreateCompoundDocument  | .NET 6.0 | .NET 6.0 |  34.76 us | 0.460 us | 0.430 us |   7.2021 |  0.8545 |   44.65 KB |
| ReadDocFile             | .NET 6.0 | .NET 6.0 | 488.34 us | 5.804 us | 5.429 us | 207.0313 | 34.1797 | 1269.61 KB |
| ReadOnePropFromADocFile | .NET 6.0 | .NET 6.0 | 491.85 us | 8.118 us | 7.593 us | 206.0547 | 34.1797 | 1267.19 KB |
| CreateCompoundDocument  | .NET 8.0 | .NET 8.0 |  30.61 us | 0.528 us | 0.494 us |   7.2021 |  0.9766 |   44.66 KB |
| ReadDocFile             | .NET 8.0 | .NET 8.0 | 503.43 us | 6.619 us | 5.867 us | 207.0313 | 40.0391 | 1269.63 KB |
| ReadOnePropFromADocFile | .NET 8.0 | .NET 8.0 | 505.28 us | 7.207 us | 6.018 us | 205.0781 | 39.0625 | 1267.22 KB |

current master branch

| Method                  | Job      | Runtime  | Mean      | Error    | StdDev   | Gen0     | Gen1    | Allocated |
|------------------------ |--------- |--------- |----------:|---------:|---------:|---------:|--------:|----------:|
| CreateCompoundDocument  | .NET 6.0 | .NET 6.0 |  37.22 us | 0.532 us | 0.779 us |   6.8359 |  0.1221 |  42.19 KB |
| ReadDocFile             | .NET 6.0 | .NET 6.0 | 474.51 us | 6.711 us | 6.277 us | 162.1094 | 24.4141 | 995.83 KB |
| ReadOnePropFromADocFile | .NET 6.0 | .NET 6.0 | 475.42 us | 8.014 us | 6.692 us | 162.1094 | 25.3906 | 993.43 KB |
| CreateCompoundDocument  | .NET 8.0 | .NET 8.0 |  33.33 us | 0.569 us | 0.504 us |   6.8359 |  0.1221 |   42.2 KB |
| ReadDocFile             | .NET 8.0 | .NET 8.0 | 403.23 us | 7.268 us | 6.798 us | 162.1094 | 27.3438 | 995.86 KB |
| ReadOnePropFromADocFile | .NET 8.0 | .NET 8.0 | 400.61 us | 7.676 us | 7.539 us | 162.1094 | 26.3672 | 993.43 KB |

The performance numbers are a tad variable when running on my laptop when the times are so low but the memory allocations have gone down quite a bit.

I'll try to give it a run over a larger number of files and for writing to make sure no issues pop out.