is the install counter on lib.haxe.org working?

[nadako]

Subj. Because e.g. for hxnodejs it shows 0 installs, though I downloaded it myself at least once.

[markknol]

@jasononeil @ncannasse bump.

What is needed to fix this? I can imaging it's very discouraging for library authors to see the download count stay the same. See http://lib.haxe.org/search/?v=hxnodejs

The popularity list also does not change because of this.

[markknol]

+1

[nadako]

I'll look into it.

[ibilon]

I launched my local haxelib server and the download counter is not 0, strange.

[ibilon]

So not all version of the library can have their counter increase, and I get a invalid field access : index if I trace the error, tacked it to https://github.com/HaxeFoundation/haxelib/blob/development/src/haxelib/server/Repo.hx#L319-L326

[markknol]

You will never notice because the function seems to be try/catched https://github.com/HaxeFoundation/haxelib/blob/development/src/haxelib/client/Main.hx#L769

[nadako]

This seems to be related to the preview/previewNum being null and passed to the SPOD query. Gotta dig into SPOD for that.

[ibilon]

after removing the try/catch ;)

For a lib without preview and previewNum: it's the preview field, commenting it prevent the crash but doesn't find the version, commenting also previewNum "works", but not on the right version.

[ibilon]

Well if you're ok with hacks you can do:

var version = SemVer.ofString(version);
var vl = Version.manager.search(
    $project == p.id &&
    $major == version.major &&
    $minor == version.minor &&
    $patch == version.patch
);

var v = null;
for (ve in vl)
    if (ve.preview == version.preview && ve.previewNum == version.previewNum)
        v = ve;

if( v == null )
    throw "No such Version : " + version;
v.downloads++;
v.update();
p.downloads++;
p.update();

it works, but that's sad to do part of a select manually.

[nadako]

It generates wrong query expression for the enum, it doesn't respect that it's nullable and tries to get Type.enumIndex(preview) which fails with the invalid field access. This may be related to 6d41c431e4d3056d4a8269c4aedc46acb408cafe

[nadako]

Well if you're ok with hacks you can do

another version is to construct the SQL manually, but I'd like to look into fixing RecordMacros first

[nadako]

SPOD is a real mess, see https://github.com/HaxeFoundation/haxe/issues/4932 and https://github.com/HaxeFoundation/haxe/issues/4931

I don't feel like messing with RecordMacros at the moment, so I'll just hand-craft the SQL here...

[nadako]

I cherry-picked the fix to the master branch which should be auto-deplyed from Travis, let's see if it works...

[nadako]

It works \o/

[markknol]

This is great! Is it online yet?

[nadako]

Yeah it's online!

[markknol]

It works, I tried with a lib that had no downloads yet, now it has 1.

[markknol]

Are installs from CI also count as download count? Can we exclude them? Otherwise it might be a bit unfair.

[nadako]

We could check for travis/appveyor/whatever env vars and skip invocation of postInstall