Code completion revealed

From the first version of Toad we have always had some form of code completion (It was originally called Toad insight or something like that) but it has always been a fairly simple affair where you either got a list of columns in select statements or a long list of objects of many different types completely unaffected by the context in which you invoke the code completion.

Starting in 4.0 the code completion support was completely rewritten to be very context aware. Now Toad should know that if you are writing a select statement and are between SELECT and FROM you probably want a list of columns and if you write JOIN you probably want to join on a foreign key from tables already specified etc. There has been little bit of a backlash on this and I have to admit that this code did have some bugs and when that happened you usually didn't end up with any code completion at all which is worse than what we had in 3.x. Hopefully for the soon to be released 4.1 release most of these should have been fixed and you will really see the benefits of the new framework.

Handling code completion in this way is kind of complicated. The main problem is you can't do strict parsing because we are not dealing with complete statements. Instead we have to do our best at guessing what the you are aiming at writing.

So let me go through the changes in more detail what new features are introduced with this framework.

Smarter auto invocation

The most obvious change that people will immediately see is that as you start typing a select statement the moment you write "FROM" and then space you will be presented with a list of tables and views. Before 4.0 the only time we automatically invoked code completion was when you typed a dot. In 4.0 we have certain "hot tokens" that invoke code completion after you write a space as well. The full list of tokens isn't that important, but if you ever think we are missing something just let us know because it is very easy for us to change it.

Context sensitive object lists

We now look at the statement type to decide what kind of object list you probably are interested. For instance if you write "CALL" or "EXEC" you are probably interested in a list of procedures, when you are writing "UPDATE" you probably want a list of tables while when you are in the "FROM" clause of a select statement you probably want both tables, views and in case it's supported functions returning tables.

Context aware sub-select completion

Toad should now keep track of sub selects so that it knows which objects you can address when you are writing complex sub-selects as shown in the screenshot below.

Auto generated joins

When you are writing select statements we now automatically generate the full join statement based on foreign keys in the database. We can generate both the part following "JOIN", the part following "ON" or in case of a cross join the part in the "WHERE" clause.

Snippet code completion

In some cases the code completion generated still need you to fill in some extra data to be a valid statement. We now do this by inserting an auto generated code snippet. Specifically we use this when you are using code completion on code to let you fill in the parameters of the call we generate.

Complete statement generation

We can now generate the full statement for some statements after you have specified the database object to operate on. Specifically this includes update, insert and delete statements.

Good handling of owners

This only applies to SQL Server, but in 4.0 we really didn't have that good support for handling owners in code completion. As of 4.1 that we now handle owners in a really clever way. Basically it should now "just work". If SQL Server accepts the syntax we should handle it correctly.

Support for local variables in batches

On SQL Server we now keep track of local variables defined in a batch and includes these in the places where they make sense.

Options available for code completion

The option page for code completion has changed quite a bit between 3.x and 4.0.

First some things that haven't changed. The auto list members, quoting, casing, expand "*" and include system objects work exactly as before so I will not go into these (And they should be pretty self explanatory).

Item sort order

This setting now has one more item which is the default and that is the "Each category". This means that the code completion list is sorted per type of items. For instance in a select statement we start with the "*" and then columns followed by variables, snippets and a bunch of other stuff. Chances are that you are looking for the columns and in this case they will be at the top instead of being lost in the list of hundreds of function snippets also included.

Merge secondary items

When you invoke code completion in some cases we have items that are common to choose from and some items that are not so common. Normally the not so common items are not included in the code completion, but if you want to get to them you can still do so by choosing the last item in the list which says (More).

If this option is checked this means that we should generate the code completion list with all items right away instead of separating this into two different lists. This is the default.

Known issues

The one known issue in regards to code completion is that we currently do not honor the setting to separate statements with double newlines (On by default I think). For version control to function property with multiple statements in the editor you need to end you statements with ';' (Except for SQL Server which doesn't have this issue).