I'm working on making a simple SQL UDF, to cast from char/varchar to time.
Since it is supposed to be something generic, I wanted the user to specify the format his input is, so I could just use something like
cast(user_time as time(0) format user_format)
but it doesn't work.
I'd like to know if it is possible to use a format as an UDF parameter and if it is, how it should be used in this cast. I could break it apart to read the format and make the time match, but I'd rather avoid it if there's a simpler way to do it.
Just to clarify, I'm using Teradata 14, and I have to use SQL, so an UDF in C is not really an option for me.
Thanks in advance
You can't pass a FORMAT as a parameter to a CAST.
You might be able to do this using TO_TIMESTAMP instead, but why do you need a UDF for that?
myUDF('12:34:45', 'hh:mi:ss')
is not much shorter than an old style Teradata CAST:
hh:mi:ss' (time(0), FORMAT 'hh:mi:ss')
Related
How do I remove the ".000000" part of the "2386.000000" field? I want to leave only the numerical part before the dot in databricks
You can use cast
select cast(2386.000000 as int) i
There are many ways to convert a float to an int. The cast() function is just one.
Please see the following link for all supported Spark SQL functions.
https://spark.apache.org/docs/2.3.0/api/sql/index.html#round
In my solution, I use the round() function. As long as you get the correct answer, the path you take may differ.
I have a TSQL view that processes multiple gigabytes of data in a SQL Server 2016 environment. In this view, there are multiple times where I am comparing if a DateTime value is before/after a static date, traditionally represented as a string literal like '2018-07-11'.
An example comparison would be:
SELECT MyId, MyValue FROM MyTable WHERE MyDate = '2018-07-11'
While looking for a way to use a DateTime literal instead of a string, I came across examples using ODBC DateTime strings like so:
SELECT MyId, MyValue FROM MyTable WHERE MyDate = {d '2018-07-11'}
When I compare the query plan I get the same result, even when I make up more advanced queries.
I started using this format in an attempt to prevent the auto-conversion of string to DateTime in queries, but I haven't been able to find any good documentation explaining any side effects of using ODBC functions. I'm not sure if this acts the same way as a string literal or if it is interpreted as a date.
If this was a UDF or Stored Procedure, I'd have the ability to declare a DateTime variable for use in the query, but in a VIEW this is not possible, nor would it be feasible because there are a lot of DateTime literals in the actual version of the query.
So in conclusion, does someone have any concrete reasons for or against using this {d '2018-07-11'} format (besides it potentially not being valid in a non SQL Server environment)?
I want to ensure that I'm not shooting myself in the foot here on a code review.
PS: I apologize for the vague examples and semi-open-ended question, I am not allowed to disclose any actual source code.
Thanks!
EDIT: I forgot to mention that I could also use DATEFROMPARTS(2018, 07, 11), but I wasn't sure if this would be looked at weirdly by the query optimizer.
The ODBC literal has the slight advantage that it can never be interpreted as YYYY-DD-MM, which is possible with one internationalization setting.
You can avoid ambiguity by using 'YYYYMMDD' format. This format is not affected by settings.
I prefer not using the ODBC, just because it seems to involve more clutter in the query. I admit to also preferring the hyphenated form (consistent with the ISO standard and other databases). But you have three alternatives. Possibly the safest for general purpose, SQL-Server-only code is the unhyphenated form.
A literal is a literal. It is transformed into a value during parsing. The value is used later.
Here is the list of DateTime literals that SQL Server supports. ODBC is a supported format.
So, if only using SQL Server then there is no difference. Different SQL flavors may reject the ODBC syntax. I do not believe it is ANSI SQL, so "less standard"?
Don't know why but while running the command
select CONVERT(datetime, getdate(),101)
gets '2016-10-27 15:53:12.743', which is the desired result.
However when the same command is run in a Stored Procedure like
if #CodeFilter2 is not null
select #CodeFilter2=CONVERT(datetime,GETDATE(),101)
yields 'Oct 27 2016 3:55PM'.
Please help me understand as to why is this happening.
Thanks in advance!
You're converting a datetime to a datetime? Are you expecting that the format used in convert sticks around? It's just ignored, since it has no meaning for the conversion you're doing. If you want to convert a datetime to a varchar, you need to use something like convert(varchar(max), getDate(), 101), which will give you the correct output 10/27/2016 - I have no idea why you'd expect either of your samples to be correct; they simply happen to work that way because the default conversion (based on locale and other context, which is very variable) happens to be your desired result (in the first case).
If you need to rely on specific formatting you must use explicit formatting. Or let the application handle it instead of the DB server. The proper format for the ODBC cannonical form (which seems to be the one you desire) is 121. Make sure you're converting to varchar or nvarchar, not datetime.
Is there a way to change the default format of a date in Postgres?
Normally when I query a Postgres database, dates come out as yyyy-mm-dd hh:mm:ss+tz, like 2011-02-21 11:30:00-05.
But one particular program the dates come out yyyy-mm-dd hh:mm:ss.s, that is, there is no time zone and it shows tenths of a second.
Apparently something is changing the default date format, but I don't know what or where. I don't think it's a server-side configuration parameter, because I can access the same database with a different program and I get the format with the timezone.
I care because it appears to be ignoring my "set timezone" calls in addition to changing the format. All times come out EST.
Additional info:
If I write "select somedate from sometable" I get the "no timezone" format. But if I write "select to_char(somedate::timestamptz, 'yyyy-mm-dd hh24:mi:ss-tz')" then timezones work as I would expect.
This really sounds to me like something is setting all timestamps to implicitly be "to_char(date::timestamp, 'yyyy-mm-dd hh24:mi:ss.m')". But I can't find anything in the documentation about how I would do this if I wanted to, nor can I find anything in the code that appears to do this. Though as I don't know what to look for, that doesn't prove much.
Never mind :'(
I found my problem. I was thinking that I was looking directly at the string coming back from the database. But I was overlooking that it was reading it as a Timestamp and then converting the Timestamp to a string. This was buried inside a function called "getString", which is what threw me off. I was thinking it was ResultSet.getString, but it was really our own function with the same name. Oops. What idiot wrote that function?! Oh, it was me ...
Thanks to all who tried to help. I'll give you each an upvote for your trouble.
I believe the table columns are specified differently. Try these variants:
timestamp
timestamp(0) no millis
timestamptz with timezone
timestamptz(0) with timezone, no millis
With which client are you running the select statements? Formatting the output is the application's responsibility, so without knowing which application you use to display the data, it's hard to tell.
Assuming you are using psql, you can change the date format using the SET command:
http://www.postgresql.org/docs/current/static/sql-set.html
Which is essentially a way to change the configuration parameters. The ones that are responsible for formatting data are documented here:
http://www.postgresql.org/docs/current/static/runtime-config-client.html#RUNTIME-CONFIG-CLIENT-FORMAT
Daniel tells me to post my findings as an answer and accept it to close the question. Okay.
I found that the date format I was seeing that did not include a time zone was not what was coming directly from Postgres, but that there were a couple of function calls that I was missing that converted the incoming date to a java.util.Timestamp, and then from the java.util.Timestamp to a String. It was in this conversion from the Timestamp to the String that the time zone was defaulting to EST.
In my own humble defense, my mistake was not as dumb as it may sound. :-0 We had the execution of the query in a subclass that read the results into a List, which we do to allow modification of the query results before output. (In this case we are adding a coule of columns that are derived from the stored columns.) Then we have a set of functions that resemble the JDBC functions to pull the data out of the List, so a calling program can easily switch from processing a query directly to processing the List. When I was wrestling with the date format problem, it just didn't register on me that I wasn't looking at "real JDBC", but at "simulated JDBC" calls.
What is the general guidance on when you should use CAST versus CONVERT? Is there any performance issues related to choosing one versus the other? Is one closer to ANSI-SQL?
CONVERT is SQL Server specific, CAST is ANSI.
CONVERT is more flexible in that you can format dates etc. Other than that, they are pretty much the same. If you don't care about the extended features, use CAST.
EDIT:
As noted by #beruic and #C-F in the comments below, there is possible loss of precision when an implicit conversion is used (that is one where you use neither CAST nor CONVERT). For further information, see CAST and CONVERT and in particular this graphic: SQL Server Data Type Conversion Chart. With this extra information, the original advice still remains the same. Use CAST where possible.
Convert has a style parameter for date to string conversions.
http://msdn.microsoft.com/en-us/library/ms187928.aspx
To expand on the above answercopied by Shakti, I have actually been able to measure a performance difference between the two functions.
I was testing performance of variations of the solution to this question and found that the standard deviation and maximum runtimes were larger when using CAST.
*Times in milliseconds, rounded to nearest 1/300th of a second as per the precision of the DateTime type
CAST is standard SQL, but CONVERT is only for the dialect T-SQL. We have a small advantage for convert in the case of datetime.
With CAST, you indicate the expression and the target type; with CONVERT, there’s a third argument representing the style for the conversion, which is supported for some conversions, like between character strings and date and time values. For example, CONVERT(DATE, '1/2/2012', 101) converts the literal character string to DATE using style 101 representing the United States standard.
Something no one seems to have noted yet is readability. Having…
CONVERT(SomeType,
SomeReallyLongExpression
+ ThatMayEvenSpan
+ MultipleLines
)
…may be easier to understand than…
CAST(SomeReallyLongExpression
+ ThatMayEvenSpan
+ MultipleLines
AS SomeType
)
CAST uses ANSI standard. In case of portability, this will work on other platforms. CONVERT is specific to sql server. But is very strong function. You can specify different styles for dates
You should also not use CAST for getting the text of a hash algorithm. CAST(HASHBYTES('...') AS VARCHAR(32)) is not the same as CONVERT(VARCHAR(32), HASHBYTES('...'), 2). Without the last parameter, the result would be the same, but not a readable text. As far as I know, You cannot specify that last parameter in CAST.