Various SQL / MS SQL Server Routines

Problem/Question/Abstract:

This is a bit of a departure from my normal style of Q & A, but I thought it would be useful for those of you who do a lot of database-related stuff.

Answer:

General Purpose Stuff

Face it, code reuse is big issue irrespective of programming in an object-oriented environment. Especially, with database application programming, the stuff you do over and over again can be a bit tedious. So, I've created a bunch of general purpose functions that I use in my code - a lot.

An On-the-Fly SELECT Statement Generator

The first function CreateSelect creates a SELECT or SELECT DISTINCT statement that can be used in whenever you create a TQuery for querying against a local table like Paradox. It's most useful for when you want to create a query from a table at runtime, but don't know the fields you want to include. For instance, I created an adhoc querying tool where users could select the fields they wanted displayed in a graph by dragging the field names from on list box into another. From there, I passed the selected fields list to CreateSelect and constructed a SELECT statement on the fly. It was a one-line call as opposed to the multiple lines I'd normally have to write for a query. Here's the code:

{==========================================================
This function will create a SELECT or SELECT DISTINCT SQL
statement given input from a TStrings descendant like a
list. It will properly format the list into field decla-
rations of a SQL statement then, using the supplied
TableNm parameter, will construct an entire statement that
can be assigned to the SQL property of a TQuery.

Params:  Distinct  SELECT DISTINCT or regular SELECT
          TableNm   Table name: Should either be a fully
                    qualified table name, or preceeded by
                    an alias (ie, ':DbName:MyTable.db')
          FieldList Any TStrings descendant will work here,
                    like the Items property of a TListBox.
==========================================================}

function CreateSelect(Distinct: Boolean;
  TableNm: string;
  const FieldList: TStrings)
  : TStrings;
var
  Sql: TStringList;
  I: Integer;
  buf,
    QueryType: string;
begin
  //First, instantiate the SQL lines list
  Sql := TStringList.Create;

  //Determine whether or no this is a regular SELECT
  //or a SELECT DISTINCT query.
  if Distinct then
    QueryType := 'SELECT '
  else
    QueryType := 'SELECT DISTINCT ';

  buf := QueryType;

  try
    //Now add the fields to the select statement
    //Notice that if we're on the last item,
    //we don't want to add a trailing comma.
    for I := 0 to (FieldList.Count - 1) do
      if (I <> FieldList.Count - 1) then
        buf := buf + FieldList[I] + ', '
      else
        buf := buf + FieldList[I];

    //Now, put the query together
    Sql.Add(buf);
    Sql.Add('FROM "' + TableNm + '"');
    Result := Sql;
  finally
    Sql.Free;
  end;
end;

Here's a code snippet from one of my programs that implements the function:

var
  qry: TQuery;
begin
  qry := TQuery.Create(nil);
  with qry do
  begin
    Active := False;
    DatabaseName := 'MyLocalDatabase';
    SQL := CreateSelect(True, ExtractFilePath(ParamStr(0))
      + 'Process.DB', lbSelectedFlds.Items);
    try
      Open;
      ....other code....
    finally
      Free;
    end;
  end;
end;

WHERE It Is....

I use a lot of TStrings types in my code because they make it easy to pass a bunch of values at once. Especially with selection criteria in queries, having the capability to select on multiple values is a boon to efficiency. The three functions below can be added to a WHERE clause in a SQL statement for multiple-value searching. All you need to do is assemble the search criteria into a TStrings type like a TStringList or use even the Items property of a TListBox to be able to use these functions. Respectively, they allow you to create a basic multiple-value criteria statement (i.e. (fldName = 'thisvalue') OR (fldName = 'thatvalue'); an IN selection criteria statement; and a multiple-value LIKE selection criteria statement. Here they are:

{Returns a '(FldName = FldValue) OR (FldName = FldValue)' etc string from
a list of values. Useful for translating TListBox values into SQL strings
IsValString is a boolean to test if the list values are a value string; that is,
it contains spaces, in which case, you would want double-quotes.}

function BuildSQLSetString(fldName: string; const List: TStrings;
  IsValString: Boolean): string;
var
  I: Integer;
  buf: string;
begin
  Result := '';
  for I := 0 to (List.Count - 1) do
  begin
    if IsValString then
      buf := '(' + fldName + ' = ''' + List[I] + ''') OR '
    else
      buf := '(' + fldName + ' = ' + List[I] + ') OR ';

    Result := Result + buf;
  end;
  Result := Copy(Result, 1, Length(Result) - 4);
end;

//This will build an IN statement

function BuildSQLINString(fldName: string; const List: TStrings;
  IsValString: Boolean): string;
var
  I: Integer;
  buf: string;
begin
  Result := '';
  for I := 0 to (List.Count - 1) do
    if IsValString then
      buf := buf + '''' + List[I] + ''', '
    else
      buf := buf + List[I] + ', ';

  buf := Copy(buf, 1, Length(TrimRight(buf)) - 1);

  Result := fldName + ' IN (' + buf + ')';
end;

//This will build a LIKE statement

function BuildSQLLikeString(fldName: string; const List: TStrings;
  IsValString: Boolean): string;
var
  I: Integer;
  buf: string;
begin
  Result := '';
  for I := 0 to (List.Count - 1) do
  begin
    if IsValString then
      buf := '(' + fldName + ' LIKE ''' + TrimRight(List[I]) + '%'') OR '
    else
      buf := '(' + fldName + ' LIKE ' + List[I] + '%) OR ';

    Result := Result + buf;
  end;
  Result := Copy(Result, 1, Length(Result) - 4);
end;

Notice that in addition to the field name and value list, there's a parameter in each function called IsValString. Set this to true if the field you're searching on is a string field. Remember, discrete string values in SQL must be delimited by single quotes.

Some MS SQL Server Stuff

A little over a year ago, I moved completely away from doing local table processing with Paradox and started doing all my processing under a Client/Server environment using MS SQL Server. I was in for a big surprise when I made the switch because many of the things that I could do with Paradox tables that I took for granted, weren't as easily done in SQL Server. Yikes! So, what I did was create some generic routines specifically for Client/Server applications to make my life much easier. Let's see what I've got...

How Many Was That?

For those of you who work with local tables, you know how easy it is to get the number of records in a table: Simply connect a TTable object to the table in question and query the RecordCount property. Well, in SQL Server, that's not so easy because the idea of a "record" is non-existent in this set-based data environment. But, just because you can't get the information from the table directly, doesn't mean that it doesn't exist. In fact, the information is stored in the system tables. Here's a function that will get you the number of rows contained in a SQL Server table:

// =============================================================================
// Returns the SQL DB path of the specified BDE alias.
// =============================================================================

function GetSQLDBPath(const AliasName: ShortString): ShortString;
var
  ParamList: TStrings; // List of parms from the BDE
  Path: ShortString; // Path returned from the BDE
begin
  Result := '';
  ParamList := TStringList.Create;
  try
    Session.GetAliasParams(AliasName, ParamList);
    Path := ParamList[0];
    Result := Copy(Path, Pos('=', Path) + 1, Length(Path) - Pos('=', Path) + 1);
  finally
    ParamList.Free;
  end;
end;

function GetSQLTableRows(SesName: string; DB: TDatabase; TableName: string): Integer;
var
  qry: TQuery;
begin
  qry := TQuery.Create(nil);
  with qry do
  begin
    Active := False;
    SessionName := sesName;
    DatabaseName := DB.DatabaseName;
    SQL.Add('SELECT DISTINCT I.Rows');
    SQL.Add('FROM ' + GetSQLDBPath(DB.DatabaseName) + '..SysIndexes AS I INNER JOIN');
    SQL.Add('     ' + GetSQLDBPath(DB.DatabaseName)
      + '..SysObjects AS O ON (I.ID = O.ID)');
    SQL.Add('WHERE (O.Type = ''U'') AND (O.Name = ''' + TableName + ''')');
    try
      Open;
      Result := FieldByName('Rows').AsInteger;
    finally
      Free;
    end;
  end;
end;

Some of you might be wondering: "Of what use is this function?" Well, for some of you, indeed, it might be of no use whatsoever. But in the type of applications I write that query datasets with records numbering in the several millions, in some steps, I only want to continue processing if my result sets aren't empty. This function gives me a quick way of checking if I need to continue or not. So there!

To Transact or Not To Transact (SQL, that is)

If you're using SQL Server, but not using Transact SQL, you're missing out on a lot of functionality. I know, I know, there are those cross-platform junkies out there that will only write the most generic code so they can deploy their DB apps on other servers. That's all well and good, but for the most of us, we only have a single server platform, and to not use its native functionality is to deprive ourselves of a lot of functionality. For instance, how many of you create temporary tables when doing a series of queries? C'mon, raise yer hands high! Using strict ANSI SQL with Delphi, you would have to create three objects: a TQuery to perform the query, a TBatchMove to move the result table, and a TTable to receive the results. Yikes! But with SQL Server, you can perform a SELECT INTO query use only a single object: a TQuery. Check out the code snippet from one of my programs below:

var
  qry: TQuery;
begin
  qry := TQuery.Create(nil);
  with qry do
  begin
    Active := False;
    SessionName := ClassSession.SessionName;
    DatabaseName := DBName;
    SQL.Add('SELECT DISTINCT T.PatientID, T.Symptom);
      SQL.Add('INTO ' + EvalTemp1);
      SQL.Add('FROM ' + SymptomTable + ' T (TABLOCK), HoldingTable H (TABLOCK)');
      SQL.Add('WHERE (H.PatientID = T.PatientID) AND (Age > 65) ');
      SQL.Add('AND (Thiazides_Now IS NULL) AND (GOUT IS NULL)');
      try
        ExecSQL;
      finally
        Free;
      end;
  end;
  // ....the rest of the code....

This simple query above accomplishes with one object using Transact-SQL that it normally takes three objects to do with standard ANSI SQL. Notice that the query is executed using ExecSQL as opposed to Open. If you're using Transact-SQL, everything happens on the server, so you get no cursors back. That might present a problem in some circumstances, but for the most part, since you're doing nothing but set logic, having a cursor to a result set isn't necessary.

Not only does Transact-SQL make it shorter to do extracts, it has some cool syntax that allows you to really short-circuit both code and performance. For instance, one of the things that has always irked me about SQL is doing a two-table update; that is, updating a field in a table from values contained in another table. In Paradox, it's simple QBE query with matching example elements. In ANSI SQL, it involves a subquery. Let's look at an example:

Standard ANSI SQL two-table update:

UPDATE ClaimsTbl
SET History = (SELECT Activity
               FROM ActivityTbl AS A
               WHERE (ClaimsTbl.Patient = A.Patient))

Not too complex, but let me tell you, it's as slow as molasses on SQL Server. On the other hand, the following Transact-SQL UPDATE statement works lickety-split!

UPDATE ClaimsTbl
FROM ClaimsTbl C, Activity A
SET C.History = A.Activity
WHERE (C.Patient = A.Patient)

Not much different from the example above, but since SQL Server is equipped to deal with this syntax in an optimized fashion, it works much faster.

Why have I spent so much time covering this stuff? Well, I'm a real proponent of creating systems that run in the most optimized fashion I can make them. And that means that I use all the tools available to me. Yes, it can be argued that I'm locking myself into a specific platform's functionality, but since I only have one platform that I'm dealing with, why not use it? I realize that I haven't covered hardly any of the Transact-SQL syntax. If you've got it, I'll leave it up to you to go hunting for the information (BTW, a great source is the SQL Server Books Online reference that comes with the software). The bottom line is this: If you're building systems that address a single server platform, make the best use of that server's capabilities.

But Wait! There's More!

Below are some general purpose routines that I use on a regular basis for accessing my SQL Server databases. Check 'em out:

//Uses Session.GetTableNames to get a list of tables from the specified database

procedure GetTableList(DBName: string; const TableList: TStrings);
var
  db: TDatabase;
begin
  db := TDatabase.Create(Application);
  with db do
  begin
    DatabaseName := DBName;
    LoginPrompt := False;
    Connected := True;
  end; { with }
  try
    Session.GetTableNames(db.DatabaseName, '', False, False, TableList);
    db.Close;
  finally
    db.Free;
  end;
end;

//Uses GetTableList to see if a particular table exists in database.

function TableExists(var ProcSession: TSession; var DB: TDatabase;
  TableName: string): Boolean;
var
  TableList: TStrings;
begin
  Result := False;
  TableList := TStringList.Create;
  try
    ProcSession.GetTableNames(DB.DatabaseName, '', False, False, TableList);
    if (TableList.IndexOf(TableName) > 0) then
      Result := True;
  finally
    TableList.Free;
  end;
end;

//Performs a series of drops for all table names contained in the input array.
//Very useful for cleaning up a bunch of temporary tables at once.

procedure CleanMSSQLTbl(var ProcSession: TSession; DBName: string;
  TableNames: array of string); overload;
var
  sqlEI: TQuery;
  I: Integer;
begin
  for I := Low(TableNames) to High(TableNames) do
  begin
    sqlEI := TQuery.Create(nil);
    with sqlEI do
    begin
      SessionName := ProcSession.SessionName;
      DatabaseName := DBName;
      with SQL do
      begin
        Add('if exists (select * from sysobjects where ' +
          'id = object_id(''' + TableNames[I] + ''') and sysstat & 0xf = 3)');
        Add('drop table ' + TableNames[I]);
        try
          ExecSQL;
        finally
          Free;
        end;
      end;
    end;
  end;
end;

//Will return a TStrings list containing the results of a query

function SQLLoadList(SesName, TblName, FldName: string): TStrings;
var
  qry: TQuery;
begin
  Result := TStringList.Create;
  qry := TQuery.Create(nil);
  with qry do
  begin
    Active := False;
    if (SesName <> '') then
      SessionName := SesName;
    DatabaseName := ExtractFilePath(TblName);
    SQL.Add('SELECT DISTINCT ' + FldName);
    SQL.Add('FROM "' + TblName + '"');
    try
      Open;
      while not EOF do
      begin
        Result.Add(FieldByName(FldName).AsString);
        Next;
      end;
    finally
      Free;
    end;
  end;
end;

Notice in most of the functions above, that they either require a TSession or a TSession.SessionName as one of their formal parameters. This is because these functions are thread-safe. In order to use these in a multi- threaded system, you need to create a TSession instance for every thread, and database access within the scope of the thread require a session name to operate under. Otherwise you'll get access violations. Not good.

Well, that's it for now. I encourage you to use these functions and concepts in your own code. Believe me, they've saved me a lot of time!