Product articles SQL Prompt SQL Code Analysis
SQL Prompt Code Analysis: Avoid using…

24 August 2018

1 Comment

Guest post

This is a guest post from Phil Factor. Phil Factor (real name withheld to protect the guilty), aka Database Mole, has 30 years of experience with database-intensive applications.

Despite having once been shouted at by a furious Bill Gates at an exhibition in the early 1980s, he has remained resolutely anonymous throughout his career.

He is a regular contributor to Simple Talk and SQLServerCentral.

24 August 2018

1 Comment

SQL Prompt Code Analysis: Avoid using the ISNUMERIC Function (E1029)

Avoid using the IsNumeric() function, because it can often lead to data type conversion errors, when importing data. On SQL Server 2012 or later, use the Try_Convert() or Try_Cast() function instead. On earlier SQL Server versions, the only way to avoid it is by using LIKE expressions.

Guest post

This is a guest post from Phil Factor. Phil Factor (real name withheld to protect the guilty), aka Database Mole, has 30 years of experience with database-intensive applications.

Despite having once been shouted at by a furious Bill Gates at an exhibition in the early 1980s, he has remained resolutely anonymous throughout his career.

He is a regular contributor to Simple Talk and SQLServerCentral.

Avoid using the IsNumeric() function, because it can often lead to data type conversion errors, when importing data. SQL Prompt Code Analysis rules include an Execution rule, E1029, which will alert you to use of this function, in your T-SQL. If you’re working on SQL Server 2012 or later, it’s much better to use the Try_Convert()or Try_Cast() function instead. On earlier SQL Server versions, the only way to avoid it is by using LIKE expressions.

The stated purpose of IsNumeric() is to “determine whether an expression is a valid numeric type“. It accepts int, numeric, float and money numbers, and will tell you whether the string or number you provided can be cast or converted into a number. The trouble is that it won’t tell you what sort of number; it will tell you that something like ',' or '$-.' or '-' or '0E0' can be converted to a number but leaves you to guess what sort.

Data type conversions when using IsNumeric()

Let’s take an obvious use for IsNumeric(). You are importing values from a log file that has a column that is supposed to be a floating point number. You want to be sure that each input row can be inserted into your import table, which has a float column. If not, you want the row to be held in a separate input table, for investigation. The CSV file is parsed, and your input includes the string ‘+’. You check with IsNumeric(), it passes the check and so you insert the row. Bang!

You’ve got an exception. The ‘+’ string can be cast into money (0.00), an int (0) but not a float.

Let’s flesh this out a bit so we can try things out. We’ll start by successfully filtering out a bad row, so that we can deal with it afterwards, by manual intervention.

Listing 1

Execute this, and you’ll find that the routine has successfully filtered out the bad value:

Now all we do is to change the word ‘bad’ in the VALUES list into ‘+’ and what happens?

You’ve been caught out. The ‘+’ symbol can be cast to money or an int, but not a float.

Replace IsNumeric with Try_Convert

The answer is to replace IsNumeric('+') with the more recent Try_Convert(). In Listing 1, leave our previously-problematic ‘+‘ symbol in place, and change the two final expressions that check to see if the values are numeric, as follows:

Listing 2

It all works fine, and we are back with a more robust input routine:

Flushed with success, you can then change the routine to check the dates and the tap number too.

Replacing IsNumeric with LIKE expressions

You haven’t got SQL Server 2012 or later? That’s a shame. You’ll need to build some bulkier LIKE expressions instead, in order to replace IsNumeric() with code that tells you more precisely whether the string can be cast as a float. To avoid adding all that complexity to our code, we’ll opt to hide it in our import table as a calculated column. For anyone stuck on SQL Server 2005, I’ll put the calculated column in a temporary table, rather than a table variable.

Listing 3

You don’t need to figure out these LIKE expressions because they are all in my blog ‘Is It Time To Stop Using IsNumeric()‘. There are also more explanation of the quirkiness of IsNumeric here: ‘IsNumeric() Broken? Only up to a point.

Conclusion

The IsNumeric() function was never much use, because you only rarely want to ask ‘can this string be cast to any numeric type?‘ You almost always want to know whether it can be cast specifically to money, int, numeric, float and so on. It is much better to use Try_Convert() or Try_Cast (the only difference is the way parameters are passed). Try_Parse() is more useful for parsing dates in other language formats.

If you understand the quirks of IsNumeric()and can accept them, then by all means use it. The worst of its quirks is that it doesn’t know about Unicode currency symbols, such as the UK Pound sign ‘₤’ and the Euro, ‘€’, so it says that they can’t be cast as money. Obviously, if you only want to know whether a string or numeric value can be cast as one or more of various the numeric types, and you aren’t worried about the fact that it doesn’t recognize most of the world’s currencies as money, then use IsNumeric(). However, for the majority of cases, Try_Convert() is far handier and reliable.

If you are stuck with an old version of SQL Server, then your best choice is to use the LIKE expressions that I’ve shown you in my blog ‘Is It Time To Stop Using IsNumeric()‘.

Share this post.

Share on FacebookShare on Google+Share on LinkedInTweet about this on Twitter
  • Andy Robertson

    Excellent article. Thanks Phil.

You may also like

  • Article

    How to use the SQL Prompt snippet placeholders for selecting and copying text

    Phil Factor demonstrates the power of placeholders in SQL Prompt Snippets, for selecting and copying text. They make it very easy to build queries by dragging-and-dropping from the SSMS object explorer, and also to save queries as reusable views or inline table valued functions.

  • Article

    SQL Prompt Hidden Gems: The SSMS Results Pane

    Every time you need to reuse the query results from SSMS, for example to populate another table, or to search for matching rows in another table, it will inevitably mean a lot of manual tweaking to the get the results into the right format. Louis Davidson uncovers three hidden Prompt gems (Open in Excel, Script as INSERT, Copy as IN clause) that can remove all this pain.

  • Article

    SQL Productivity with SQL Prompt

    There is an old joke that upgrading to the latest SQL Server is wasted on some DBAs, because they will still stick mainly to what worked in SQL Server 2005. This type of DBA is becoming rare, in my experience, but there is still some truth in the idea that many of us don’t get

  • Article

    Driving up database coding standards using SQL Prompt

    A strategic view of how a development team can use SQL Prompt to establish and share coding standards, through code analysis rules, formatting styles and code snippets.

  • Forums

    SQL Prompt Forum

    Write, format, and refactor SQL effortlessly