Add doc for issue #8420

[ihipop]

#8420 (comment)

Q	A
Is bugfix?	no
New feature?	no
Breaks BC?	no
Tests pass?	yes
Fixed issues	#8420

[samdark]

@ihipop, @tom--, @beowulfenator please check docs in this pull request.

[beowulfenator]

How do I add something to this PR?

[samdark]

You can't since you don't have write access to this repo. Either comment inline or, alternatively, access could be given.

[ihipop]

@samdark Thank you for your great work.
But I have some different advice.

The table may remain locked by MySQL and cannot be written to by other queries. Thus, new connection should be created for this purpose.

This is not the purpose that we establish a new connection to MySQL server, we establish a new connection to MySQL server because :

Unless whole data set has been retrieved, no other query could be done through the same connection

Even we establish a new connection to MySQL server ,tables may still remain locked by MySQL and cannot be written to by other queries ，Because MySQL keeps locks on data in the case of MyISAM on the entire table ,and In Unbuffered Query Mode, because the loop done by PHP is much more slower than Buffered Query done by PDO (Not only because PHP is slower than C, but also the developer usually do some business logic in the loop for their specify reason), the lock time to table will much more longer than Buffered Query , these two reasons are too different.

So, The lock table statement should be moving to the warning part as I've commited.

Unbuffered Query is not a silver bullet indeed, just as what @SamMousa have said:

As I've said before, no solution is valid for all cases and therefore we should not just implement something. Instead implement it yourself as a custom function or behavior in your query class.

Maybe the total solution should be done by the developer in their specify business code, for us , a well document is on call.

@beowulfenator

[ihipop]

And, also, I think that the paragraph about the memory strategy of PHP before 5.6 (which use mysqlnd) should not be removed
It explains that why batch() and each() used to seems to work (will not hit the memory limit problem, but will eat the server's memory at all) and lead a meditate to Bunffered Query

Note:
When using libmysqlclient as library PHP's memory limit won't count the memory used for result sets unless the data is fetched into PHP variables. With mysqlnd the memory accounted for will include the full result set.

Describe as: http://php.net/manual/en/mysqlinfo.concepts.buffering.php

[beowulfenator]

Can you guys give me access to the repo?

[ihipop]

@beowulfenator https://github.com/ihipop/yii2/invitations invited

[beowulfenator]

Does this look better?

[samdark]

Looks great to me.

[samdark]

@ihipop how do you find it now?

[ihipop]

@beowulfenator excellent work.
I would like to tweak a few words

[samdark]

Merged. Thank you!

[tom--]

This is good. Thanks