Records cannot be added to an empty ActiveJSON datastore

[davidstosik]

Here's the simplest way to reproduce the error:

# test.rb

require "active_hash"

class Test < ActiveJSON::Base; end

Test.create()

$ echo '[]' > tests.json
$ ruby test.rb
Traceback (most recent call last):
        3: from test.rb:7:in `<main>'
        2: from /Users/david-stosik/.rbenv/versions/2.6.5/lib/ruby/gems/2.6.0/gems/active_hash-3.0.0/lib/active_hash/base.rb:174:in `create'
        1: from /Users/david-stosik/.rbenv/versions/2.6.5/lib/ruby/gems/2.6.0/gems/active_hash-3.0.0/lib/active_hash/base.rb:497:in `save'
/Users/david-stosik/.rbenv/versions/2.6.5/lib/ruby/gems/2.6.0/gems/active_hash-3.0.0/lib/active_hash/base.rb:135:in `insert': undefined method `length' for nil:NilClass (NoMethodError)

In some scenarios, @records can be nil when checking its length and appending a new record.

This PR adds a test case that reproduces the error above, and fixes it.

My first approach was to simply replace @records = nil with @records = [], but then I figured out I could do better and improve the code base a bit, avoiding in the process duplication of code such as @records || [].

All specs still pass.

Please let me know what you think.

[kbrock]

this looks nice.

consolidates the delete_all details.
not sure that records buys us that much but good enough

could you rebase this to fixup your spec and then we can get this merged

[davidstosik]

Sorry for the delay, this is rebased now. 👌

[davidstosik]

This still seems to be a problem, so I rebased my branch.

[kbrock]

Darn. this is still outstanding. After we merge #268 I'd like to get this in.
Please ping me if I haven't gotten this in by 5/15/2023

[kbrock]

I thought we had a whole discussion here.

I view active_hash as a static data store. And if you want to treat it as a writable database, there are other database projects that may be better suited (like sqlite).

sqlite does support sqlite3 dbfile.sqlite ".dump" which will output the results as text. This may meet your needs.

[flavorjones]

@kbrock I agree. We did discuss this a bit at #234 and I think we should decline to address writing behaviors.