public / sequelize

Allows us to define our own timestamp columns.

Enum values can now be null

If the column allows null values, the validation for that column will be skipped
if the vale is null.

Closes https://github.com/sequelize/sequelize/issues/839

This commit allows the user to predefine their own timestamp columns which is
useful for databases that already exist. This also allows the user to have their
own naming conventions for the timestamp columns.

The following options have been added when defining a new table:

* createdAt (defaults to "createdAt")
* updatedAt (defaults to "updatedAt")
* deletedAt (defaults to "deletedAt")
* touchedAt (defaults to "touchedAt")

Note: The underscored option still remains useful since it's just a simple one
line option which allows flexibility for those rare cases in which people decide
to change the structure of their database from camelCase to under_score or
vice-versa.

Conflicts:
	lib/dialects/mysql/query-generator.js
	lib/query-interface.js
	lib/utils.js
	spec/dao-factory.spec.js

…andartsnl/sequelize into tandartsnl-bugfix/getterSetterPropertyInitialization

…urango-update-validations

…to janmeier-eagerloadedAttributes

Updated validator to 1.5.0 and chai-datetime to 1.1.1

Postgres was having trouble restablishing connections for non-pool clients closes #832

… times might be off slightly by a second or so when doing a comparison.

Postgres' connector-manager was only checking to see if this.client was null. We
should have been checking for "readyForQuery" status as well. We now make
this.client null if the connection has an error rather than just maintaining
stale information. This is to ensure that we retry our connection until we can
re-establish a firm connection.

This resolves issue #832 https://github.com/sequelize/sequelize/issues/832

Move the validation code to happen before the save funtion checks for enums being out of range.  This way the enum validators can catch the error and raise the event instead of throwing an error that breaks execution.