Type Definition for HDF5 module about hdf5.node HOT 24 CLOSED

I'm not familiar with what this is:-)

Would it be best to be a part of this code or separate in https://github.com/DefinitelyTyped/DefinitelyTyped?

I'm guessing this adds an interface for typescript calls?

from hdf5.node.

I guess its better to be part of this module, because of version mismatches.
I'm currently writing the type definition and could make another pull request in the next few days.

It's only an extra definition file of the javascript interface. So it doesn't change existing code.

from hdf5.node.

Right. And this adds the potential of typescript users liking this project?

Is there anything you need changed to make it nice?

from hdf5.node.

Yes, absolutly.

There is one thing.
The properties of the buffer like rank, row, and type would be better to be placed in the options argument. Then these properties would be set to required in typescript, so no one can miss to set these properties. Then the buffer is a "native" object without modified properties. But this could be a breaking change for existing projects? Or maybe enable both ways to define these properties?

from hdf5.node.

Oh I've been wanting to move to that style with the options! Then properties are clear of any reserved names. And it would fit javascrpt/JSON coding styles.

Been thinking to make the legacy of reserved rank row, and type properties only work if no options are defined.

from hdf5.node.

In other words this can be made seamless for prior users without breaking code they already have

from hdf5.node.

I'll work at changing to the options args for rank rows type this evening. You morning seems earlier than mine:-)

from hdf5.node.

Legacy must also be available when someone defines chunkSize or compression in options but use the old buffer options.

from hdf5.node.

Yep. Are you fine with getting options back as done in https://github.com/HDF-NI/hdf5.node/blob/master/test/test_h5lt.js#L136 ?

const readBuffer=h5lt.readDataset(group.id, 'Quadruple Rank', function(options) {
                    JSON.stringify(options).should.equal('{"rank":4,"endian":0}')
                });

from hdf5.node.

Yes, that's fine.

from hdf5.node.

All you're doing is really strengthening this project! Thank you

from hdf5.node.

Thank you.

But you are one who created this great module and housekeeping the code.
Thanks too.

from hdf5.node.

Maybe when we now changing the options, dims should be an array like: dims[2, 4, 5, 6] and maxDims[-1, 4,5,6];
This could make the code a little bit cleaner.
Then maybe remove the legacy properties and make a new version 0.4.0, so no user get the new spec without action.
And add a hint to the readme about the new function call.

from hdf5.node.

Yes, a release at 0.3.4 so that bug fixes are available to anyone with legacy style, tag it so a branch can be made if necessary; then make it all with the new spec.

Also I'm seeing a lot of compile warnings now with nodejs v10.1.x which I want to investigate.

Docs will get an update including the tutorials. I got some of the way thru moving to options last night but first found some other bugs.

from hdf5.node.

Should have it testable later today with {rank: 2, rows: 10, columns: 10 ....} style

sutoiku most likely is working on nodejs 10.1.x changes in the v8

from hdf5.node.

What should be documented for typescript developers?

from hdf5.node.

Maybe add an example

import { H5Type, Access, Interlace, H5OType, HLType, H5SType } from 'hdf5/lib/globals';
import { hdf5, h5lt, Hdf5Buffer } from 'hdf5';

const file: hdf5.File = new hdf5.File('test2.h5', Access.ACC_TRUNC);
const buffer: Hdf5Buffer = Buffer.alloc(8 * 10 * 8);
buffer.rank = 2;
buffer.rows = 8;
buffer.columns = 10;
buffer.type = H5Type.H5T_NATIVE_DOUBLE;

for (let j: number = 0; j < buffer.columns; j++) {
    for (let i: number = 0; i < buffer.rows; i++) {
        if (j < (buffer.columns / 2)) {
            buffer.writeDoubleLE(1.0, 8 * (i * buffer.columns + j));
        }
        else {
            buffer.writeDoubleLE(2.0, 8 * (i * buffer.columns + j));
        }
    }
}
h5lt.makeDataset(file.id, 'Waldo', buffer, { maxRows: H5SType.H5S_UNLIMITED, chunkSize: 10, compression: 7 });
file.close();

and add a hint that not all functions are currently described in typescript.

from hdf5.node.

Ah good

May I use your full name in credits?

from hdf5.node.

Yes and thanks.

from hdf5.node.

btw the current waldo test and tutorial has been changed to use the options way instead of the reserved properties

You want docs on the optional maxRows: H5SType.H5S_UNLIMITED capability right? Will add.

I was asking do typescript developers know what to do? How to make developers become aware of your new dts files? I'll learn a bit to help maintain

from hdf5.node.

Great.

Should we add directly the maxDims: [-1, 100] option?

I develop in vs code, and the d.ts files are automatically recognize, so the original Javascript functions get a description and parameters get description.

from hdf5.node.

The d.ts files are like an overlay about your existing Javascript functions. So the Javascript and dts files must match, otherwise there will be Javascript errors like: undefined is not a function.

from hdf5.node.

maxDims as an array? Might be what users gravitate towards. Will look into it

from hdf5.node.

That's the same as you mentioned in #42 higher rank datasets.

from hdf5.node.