Wrong quote position writing condensed flow

[GabrieleC]

This document:

key1: val1
key2: val2

dumped with options { flowLevel: 0, condenseFlow: true } produces output with misplaced quote:

{"key1":val1", key2":val2}

[puzrin]

Dupe of #412

[GabrieleC]

I made a simple bugfix for this, should i discard it or make a PR?

[puzrin]

I guess those options are mutually exclusive. condenseFlow was added for concrete case - URL encoding, and i would not like to spread it wide without need. If you wish this to be changed, this should be discussed first.

[GabrieleC]

When using condenseFlow: true:

• if flowLevel == -1 (default value) all collections will be write as blocks, so condenseFlow: true will be useless.
• if flowLevel >= 0 there will be the bug.

So either "condenseFlow" is useless or it's broken.

There's something I'm misunderstanding?

[puzrin]

We do nod add features "just for fun". As i said, this option was introduced for concrete use case, and not expected to be used anyhow else. If you wish this to be changed - describe your use case first, before any coding.

[GabrieleC]

The use case I'm referring is URL encoding. condenseFlow is intended to be used with flowLevel = 0, otherwise it would be useless. An example is shown in the original issue: #346.

The bug is in dumper.js, line 572, where comma is appended after the quote instead of before.

I understand this library isn't intended to be used as minifier, so if this fix isn't well accepted I will not go further.

[puzrin]

Probably i missunderstood you. Take a look at https://github.com/nodeca/js-yaml/pull/358/files, there are some tests. If something is missed and works not as intended in original PR, fix will be accepted.

Just make sure that's really a bug, because this option should be actively used by PR authors and they did not reported problem before.

[GabrieleC]

The original test doesn't throw error because it uses a collection with only one item, so no comma is written.

I think this bug wasn't included in the original PR (#358), because code is much different from the current version. Maybe PR authors never upgraded to a newer version with the bug.

I made a PR including the fix and an improvement to the original test to intercept the bug: #527