Is if...exists supported in Snowflake stored procedures? - stored-procedures

I have written a stored procedure which will be called from python. The stored procedure needs to insert the variant data into my table if the id doesn't exist or update the existing variant data where there is a match for the id. The id will be passed the way the variant data is, but for now I am just trying to get it working with a hardcoded id. The stored procedure gets called successfully from python, but then nothing gets inserted or updated in the stored procedure and the stored procedure doesn't give me an error. I am not sure if I am doing something wrong or the...
if exists (select * from my_database_table where my_variant_data:id::varchar = '123456')
... part is being ignored because it isn't supported. I haven't been able to find anything in the documentation to prove or disprove this. Does anyone know?
create or replace procedure my_stored_procedure("variant_data" variant)
returns string
language javascript
strict
execute as owner
as
$$
var insert_update_query = `
if exists (select * from my_database_table where my_variant_data:id::varchar = '123456')
begin
update my_database_table SET my_variant_data = parse_json(:1)) WHERE my_variant_data:id::varchar = '123456'
end
else
begin
insert into my_database_table(my_variant_data) select (parse_json(:1));
end
`
var result = "";
try {
var sql_insert_update_query = snowflake.createStatement({
sqlText: insert_update_query
});
var insert_update_query_result = sql_insert_update_query.execute();
result += "\n Query succeeded";
} catch (err) {
result += "\n Query failed failed: " + err.code + "\n State: " + err.state;
result += "\n Message: " + err.message;
result += "\n Stack Trace:\n" + err.stackTraceTxt;
}
return result;
$$
;
I have tested the insert and update parts of the query in the stored procedure individually and they work fine.
Insert - works as expected.
create or replace procedure my_stored_procedure("variant_data" variant)
returns string
language javascript
strict
execute as owner
as
$$
var sql_command = "insert into my_database_table(my_variant_data) select (parse_json(:1));";
var sql = snowflake.createStatement( {sqlText: sql_command, binds:[JSON.stringify(variant_data)]});
var resultSet = sql.execute();
return sql_command;
$$
;
Update - works as expected.
create or replace procedure my_stored_procedure("variant_data" variant)
returns string
language javascript
strict
execute as owner
as
$$
var sql_command = "UPDATE my_database_table SET my_variant_data = parse_json(:1)) WHERE my_variant_data:id::varchar = '123456'";
var sql = snowflake.createStatement( {sqlText: sql_command, binds:[JSON.stringify(variant_data)]});
var resultSet = sql.execute();
$$
;

Given the CODE executed needs to be valid runs on the console SQL, which this if is not, and it is fundamentally a MERGE command I would suggest flipping the code into a MERGE:
MERGE INTO my_database_table USING my_database_table
ON my_variant_data:id::varchar = '123456'
WHEN MATCHED THEN UPDATE SET my_variant_data = parse_json(:1))
WHEN NOT MATCHED THEN INSERT (my_variant_data) VALUES (parse_json(:1));
otherwise if you are want it in SP space, then I would be inclinded to break the code into a SELECT x INTO varaible FROM blar pattern and then have the IF be in SP and pick between the two blocks of SQL to run. But given it's just a merge, I would again still, do a merge.

Related

Executing sql from a javascipt UDF

I have a snowflake table being used to store records of sql being executed by a stored procedure and any error messages. The records in this table are being saved as a string with special chars escaped with javascripts escape('log') function. I then want to create a view to this log table that will return the records in an easily readable format.
My first attempt at this was to create an additional stored procedure to query the log table, pass the record to the unescape() function, then return it. Calling this procedure works as intended but we can't then create a view of this data say with something like
create view log_view as
select (call UNESCAPE_PROC());
The other idea was to use a UDF rather than a stored procedure. However this also fails as we can't execute sql code with a javascript UDF. This post touches on this idea.
My question is how can I record these executed statements in a table in such a way as they can be later viewed in a plain text readable format. I've outlined my attempts below but perhaps my approach is flawed, open to any suggestions.
Minimal working example below
Sample log table and procedure to write a statement to said table
create or replace table event_table(
event varchar,
event_stamp timestamp
);
create or replace procedure insert_to_event(stamp string)
RETURNS VARCHAR
LANGUAGE JAVASCRIPT
COMMENT = 'SP to log an event message with timestamp to event_table'
EXECUTE AS CALLER
AS
$$
// some variables to log in our event table
var str_stamp = (new Date()).toISOString();
to_log = `insert into dummy_table values(2, 'Bill', `+STAMP+`);`;
sql =
`INSERT INTO event_table (
event,
event_stamp
)
VALUES
('`+escape(to_log)+`', to_timestamp('`+str_stamp+`'));`;
var stmnt = snowflake.createStatement({ sqlText: sql });
stmnt.execute();
return "logged: "+ escape(to_log)
$$;
call insert_to_event(current_timestamp());
select * from event_table;
Stored procedure to return readable log records
CREATE OR REPLACE PROCEDURE UNESCAPE_PROC()
RETURNS VARCHAR
LANGUAGE JAVASCRIPT
COMMENT = 'SP will select a chosen column from event_table table, pass it to javascripts unescape() fn and return it'
EXECUTE AS CALLER
AS
$$
unescape_sql =
`select event from event_table`
var errs_res = [];
try {
all_logs = snowflake.execute(
{ sqlText: unescape_sql }
);
// iterate over all columns
while (all_logs.next()) {
errs_res.push(all_logs.getColumnValue(1));
}
return unescape(errs_res)
}
catch(err){
return "something went wrong: " + err
}
$$;
call UNESCAPE_PROC();
Which returns the records in a readable form as expected
However this of course wont work as part of a view eg.
-- define a view calling this procedure??
create view log_view as
select (call UNESCAPE_PROC());
Javascript user defined function can be used in a view like this, however it cannot be used to execute sql as in the stored procedures
-- use a UDF instead
CREATE OR REPLACE FUNCTION UNESCAPE_UDF()
RETURNS string
LANGUAGE JAVASCRIPT
AS
$$
unescape_sql =
`select event from event_table`
var errs_res = [];
try {
all_logs = snowflake.execute(
{ sqlText: unescape_sql }
);
// iterate over all columns
while (all_logs.next()) {
errs_res.push(all_logs.getColumnValue(1));
}
return unescape(errs_res)
}
catch(err){
return "something went wrong: " + err
}
$$
;
select UNESCAPE_UDF();
Stored procedures will solve one half of my problem for me, whilst UDF's will solve the other half. How can I combine the functionality of these two methods to solve this issue?
A much cleaner approach using parameters binding:
create or replace procedure insert_to_event(stamp string)
RETURNS VARCHAR
LANGUAGE JAVASCRIPT
COMMENT = 'SP to log an event message with timestamp to event_table'
EXECUTE AS CALLER
AS
$$
// some variables to log in our event table
var str_stamp = (new Date()).toISOString();
to_log = `insert into dummy_table values(2, 'Bill', '${str_stamp}');`;
sql = `INSERT INTO event_table (event,event_stamp)
VALUES(?, try_to_timestamp(?));`;
var stmnt = snowflake.createStatement({sqlText: sql, binds:[to_log, str_stamp]});
stmnt.execute();
return "logged: "+ to_log
$$;
Call:
call insert_to_event(current_timestamp());
-- logged: insert into dummy_table values(2, 'Bill', '2022-02-03T17:45:44.140Z');
select * from event_table;
Found a solution/workaround.
Rather than using javascripts escape/unescape functions to remove special chars from the logs, we use a regex replace eg.
create or replace procedure insert_to_event(stamp string)
RETURNS VARCHAR
LANGUAGE JAVASCRIPT
COMMENT = 'SP to log an event message with timestamp to event_table'
EXECUTE AS CALLER
AS
$$
// some variables to log in our event table
var str_stamp = (new Date()).toISOString();
to_log = `insert into dummy_table values(2, 'Bill', `+STAMP+`);`;
to_log = to_log.replace(/[`~!##$%^&*|+=?'"<>\{\}\[\]\\\/]/gi, '');
sql =
`INSERT INTO event_table (
event,
event_stamp
)
VALUES
('`+to_log+`', to_timestamp('`+str_stamp+`'));`;
var stmnt = snowflake.createStatement({ sqlText: sql });
stmnt.execute();
return "logged: "+ to_log
$$;
call insert_to_event(current_timestamp());
select * from event_table;
Which writes to the log table in an easily readable format with no need for additional stored procedures/UDF's.

Execution error in stored procedure : Unsupported type for binding argument 2undefined

I am executing
CALL LOAD_CUSTDIM_HOUSTONS:
This does the below:
I am running this code to get unique Customer Details such as Consumer No and Customer Name from STG_HOUSTON table and Get Max(CUST_DImKEY) from the DIM_CUSTOMER Key for the CUSTOMER Name from above i.e Houston's.
Next I am checking if the Max key value is null i mean no match for the CUSTOMER so Key value first time to be always 1 next time it will be max value + 1.
Once we get this bind this value and the select where clause will get the unique customer and insert into Customer Dimension table.
Stored procedure looks like this:
create or replace procedure LOAD_CUSTDIM_HOUSTONS()
returns varchar not null
language javascript
as
$$
var return_value = "";
try
{
var SQL_SMT = "SELECT DISTINCT CONSUMER_NUMBER,CUSTOMER_NAME FROM STG_HOUSTONS ORDER BY CONSUMER_NUMBER ASC";
var stmt1 = snowflake.createStatement(
{
sqlText: SQL_SMT
}
);
/* Creates result set */
var res1 = stmt1.execute();
while (res1.next())
{
var var_ConsumNo=res1.getColumnValue(1);
var var_custname=res1.getColumnValue(2);
return_value =" Inside While";
var_custname = "'" + var_custname + "'";
var t_SQLstmt = " SELECT GET_CUSTDIMKEYS(?)" ;
var t_stmt = snowflake.createStatement ({sqlText: t_SQLstmt,binds: [var_custname]});
var res2 = t_stmt.execute();
res2.next();
var_Custkey =res2.getColumnValue(1)
return_value= return_value + var_Custkey;
if (var_CustKey == null)
{
var_CustdimKey =1
}
else
{
var_CustdimKey=var_CustKey+1
}
return_value= return_value + var_CustdimKey.toString();
var SQL_INSDIMCUS="INSERT INTO DIM_CUSTOMER(CUSTOMER_KEY,CUSTOMER_ID,CUSTOMER_NAME,CONSUMER_ID,CONSUMER_NAME,CUSTOMER_IND,TODATE)" + " SELECT DISTINCT :1, CUSTOMER_NUMBER,CUSTOMER_NAME,CONSUMER_NUMBER,CONSUMER_SHPPNG_FNAME + ' ' + CONSUMER_SHPPNG_LNAME AS CONSUMER_NAME, FALSE, NULL FROM STG_HOUSTONS WHERE CONSUMER_NUMBER = :2 ;";
var stmt = snowflake.createStatement({sqlText: SQL_INSDIMCUS,binds: [var_CustdimKey,var_ConsumNo]} );
var res3 = stmt.execute();
result = "COMPLETED SUCCESSFULLY!";
}
}
catch (err)
{ result = "Failed!";
var time_st = snowflake.execute( {sqlText: `SELECT CURRENT_TIMESTAMP;`} );
snowflake.execute({
sqlText: `insert into PROC_ERROR_LOG
VALUES (?,?,?,?,?)`
,binds: ['LOAD CUSTOMER DIMENSION HOUSTONS',err.code, err.state,err.message, err.stackTraceTxt]
});
}
return return_value;
$$
;
This is the function I am using it to get max(DIM CUST KEY).
create or replace function GET_CUSTDIMKEYS ( CUSTOMERNAME varchar )
returns bigint
as 'Select max(CUSTOMER_KEY)
from PUBLIC.DIM_CUSTOMER
where CUSTOMER_NAME = CUSTOMERNAME ';
I tried with Binding variable / direct substitution of the values to dynamically create the SQL and execute it.
Using DBeaver21.1.1 IDE to create snowflake stored procedure and functions
This is the execution log Created by the IDE.
2021-10-05 20:44:48.030 - SQL Error [100183] [P0000]: Execution error in store procedure LOAD_CUSTDIM_HOUSTONS:
Unsupported type for binding argument 2undefined
At Snowflake.execute, line 49 position 14
org.jkiss.dbeaver.model.sql.DBSQLException: SQL Error [100183] [P0000]: Execution error in store procedure LOAD_CUSTDIM_HOUSTONS:
Unsupported type for binding argument 2undefined
At Snowflake.execute, line 49 position 14
at org.jkiss.dbeaver.model.impl.jdbc.exec.JDBCPreparedStatementImpl.executeStatement(JDBCPreparedStatementImpl.java:208)
at org.jkiss.dbeaver.ui.editors.sql.execute.SQLQueryJob.executeStatement(SQLQueryJob.java:510)
at org.jkiss.dbeaver.ui.editors.sql.execute.SQLQueryJob.lambda$0(SQLQueryJob.java:441)
at org.jkiss.dbeaver.model.exec.DBExecUtils.tryExecuteRecover(DBExecUtils.java:171)
at org.jkiss.dbeaver.ui.editors.sql.execute.SQLQueryJob.executeSingleQuery(SQLQueryJob.java:428)
at org.jkiss.dbeaver.ui.editors.sql.execute.SQLQueryJob.extractData(SQLQueryJob.java:813)
at org.jkiss.dbeaver.ui.editors.sql.SQLEditor$QueryResultsContainer.readData(SQLEditor.java:3280)
at org.jkiss.dbeaver.ui.controls.resultset.ResultSetJobDataRead.lambda$0(ResultSetJobDataRead.java:118)
at org.jkiss.dbeaver.model.exec.DBExecUtils.tryExecuteRecover(DBExecUtils.java:171)
at org.jkiss.dbeaver.ui.controls.resultset.ResultSetJobDataRead.run(ResultSetJobDataRead.java:116)
at org.jkiss.dbeaver.ui.controls.resultset.ResultSetViewer$ResultSetDataPumpJob.run(ResultSetViewer.java:4624)
at org.jkiss.dbeaver.model.runtime.AbstractJob.run(AbstractJob.java:105)
at org.eclipse.core.internal.jobs.Worker.run(Worker.java:63)
Line 49 is:
sqlText: `insert into PROC_ERROR_LOG
VALUES (?,?,?,?,?)`
,binds: ['LOAD CUSTOMER DIMENSION HOUSTONS',err.code, err.state,err.message, err.stackTraceTxt]
The error message:
Unsupported type for binding argument 2undefined
means the variable passed into the binding argument is not defined.
This tells me that the variable "err" was not initialized properly. I have seen some cases that under certain conditions, like when calling a function that does not exist on an object, though exception is thrown, the argument passed into the catch block is still NULL.
You need to look into the query history for this particular session, and find out which one of the queries that ran from the SP failed, and find out why.
And in the catch block, try to check the "err" object before using its attributes, something like:
if (err) {
snowflake.execute({
sqlText: `insert into PROC_ERROR_LOG
VALUES (?,?,?,?,?)`
,binds: ['LOAD CUSTOMER DIMENSION HOUSTONS',err.code, err.state,err.message, err.stackTraceTxt]
});
} else {
snowflake.execute({
sqlText: `insert into PROC_ERROR_LOG
VALUES (?,?,?,?,?)`
,binds: ['LOAD CUSTOMER DIMENSION HOUSTONS','my code', 'my state','my error', 'my stack']
});
}
This can avoid such failure.
This is in addition to Eric's answer. You should not wrap strings in single quotes when used as bind variables:
var_custname = "'" + var_custname + "'";
var t_SQLstmt = " SELECT GET_CUSTDIMKEYS(?)" ;
var t_stmt = snowflake.createStatement ({sqlText: t_SQLstmt,binds: [var_custname]});
This should be:
//var_custname = "'" + var_custname + "'";
var t_SQLstmt = " SELECT GET_CUSTDIMKEYS(?)" ;
var t_stmt = snowflake.createStatement ({sqlText: t_SQLstmt,binds: [var_custname]});
If you wrap a string type bind variable in single quotes, the single quotes become part of the variable. The practical effect is that instead of running a statement like this:
SELECT GET_CUSTDIMKEYS('MyCustomer'); -- Looks for MyCustomer
You run a statement like this:
SELECT GET_CUSTDIMKEYS('\'MyCustomer\''); -- Looks for 'MyCustomer'
The issue is first the second result set should have a while loop which was missing,
And rest of the code proceeds will solve the issue and also we have to make sure the sql statement variables, resultset variables with in the procedures are not named the same or used multiple times. And also declare or initialize variables used in the stored procedure
lessons learnt by me and happy learning.
var var_Custkey=0;
while (res2.next())
{
var_Custkey =res2.getColumnValue(1)
return_value= return_value + var_Custkey;
if (var_CustKey == 0)
{
var_CustdimKey =1
}
else
{
var_CustdimKey=var_CustKey+1
}
return_value= return_value + var_CustdimKey.toString();
var SQL_INSDIMCUS="INSERT INTO
DIM_CUSTOMER(CUSTOMER_KEY,CUSTOMER_ID,CUSTOMER_NAME,CONSUMER_ID,CONSUMER_NAME,CUSTOMER_IND,TODATE)" + " SELECT DISTINCT :1, CUSTOMER_NUMBER,CUSTOMER_NAME,CONSUMER_NUMBER,CONSUMER_SHPPNG_FNAME + ' ' + CONSUMER_SHPPNG_LNAME AS CONSUMER_NAME, FALSE, NULL FROM STG_HOUSTONS WHERE CONSUMER_NUMBER = :2 ;";
var stmt = snowflake.createStatement({sqlText: SQL_INSDIMCUS,binds: [var_CustdimKey,var_ConsumNo]} );
var res3 = stmt.execute();
}

UDF/Procedure with control flow that returns a table

I am kinda stuck with the limitations of UDFs/stored Procedures in Snowflake and wonder what I am missing. I feel like there has to be a solution, but I can't see it right now, so maybe som new input from stackoverflow can help me.
I want something (UDF/Procedure) that has an input and returns a table. I created a stored procedure that generates the SQl-Statement. But in the stored procedure I can only return a single value and no table.
My stored procedure is the following:
CREATE OR REPLACE PROCEDURE DWH.TEMP.getAccounts(NODE_NAME VARCHAR)
RETURNS VARCHAR
LANGUAGE JAVASCRIPT
EXECUTE AS OWNER
AS '
var maxLevel = 0;
function executeSQL(SQL ) {
// Prepare statement.
var stmt = snowflake.createStatement(
{
sqlText: SQL
}
);
// Execute Statement
var res = stmt.execute();
res.next();
return res.getColumnValue(1);
}
maxLevel= executeSQL(" SELECT MAX(LEVEL) as Max_Level FROM DWH.MART.VDIM_ACCOUNT");
var sqlString = "SELECT DISTINCT AccountID FROM DWH.MART.VDIM_ACCOUNT WHERE ";
var sqlWhere = "";
for (var i = 0;i<=maxLevel;i++)
{
sqlWhere += "Group_L"+ i +" in (\'"+NODE_NAME+"\')";
if(i<maxLevel)
{
sqlWhere += " OR ";
}
}
sqlString += sqlWhere;
return sqlString;
'
If I execute this with a node, for example with call GETACCOUNTS('Test') I get the result:
SELECT DISTINCT AccountID
FROM dwh.mart.vdim_account
WHERE group_l0 IN ( 'TEST' )
OR group_l1 IN ( 'TEST' )
OR group_l2 IN ( 'TEST' )
OR group_l3 IN ( 'TEST' )
OR group_l4 IN ( 'TEST' )
OR group_l5 IN ( 'TEST' )
But I want the result of above SQL-Statement returnes as a table or list. How do I do this in Snowflake?
I am stuck because I need the for loop to dynamically generate my SQL-Statement. I can only use loops in JS UDFs, but I can only execute SQL in SQL UDFs.

Trying to pass parameter as binding variable in snowflake statement

Below is my stored procedure, I'm not sure as to why it keeps throwing an error. The error I get is
SQL compilation error: syntax error line XX at position XX unexpected '?'.
I have followed the documentation here but it does not seem to work for me.
This is what I have:
CREATE OR REPLACE PROCEDURE spExample(INPUT_TABLE VARCHAR)
RETURNS VARCHAR
LANGUAGE JAVASCRIPT
AS
$$
result = "";
try {
var sql_cmd = "SELECT * FROM ?;";
var sql_stmt = snowflake.createStatement({sqlText: sql_cmd, binds:[INPUT_TABLE]});
sql_stmt.execute();
} catch(err) {
result += "Message: " + err.message;
}
return result;
$$;
Have I made a mistake somewhere?
Above answers subject your code to sql injection attack. And of course you can bind a table name to a variable in snowflake.
Do
var sql_cmd = "SELECT * FROM IDENTIFIER('?');";
It seems I had the exact same misunderstanding as the OP. It was good to find this answer.
In any case, it's a lot more flexible & more readable to use JavaScript template literals using backticks (instead of using single quotes or double quotes). They allow you to use expression interpolation in the format of
`Some text here. ${expression} Some more text here.`
Just fill in with your variable or variables (or expression).
Here is what I tried and it executed perfectly:
CREATE OR REPLACE PROCEDURE spExample(INPUT_TABLE VARCHAR)
RETURNS VARCHAR
LANGUAGE JAVASCRIPT
AS
$$
result = "";
try {
var sql_cmd = "SELECT * FROM IDENTIFIER(?);";
var sql_stmt = snowflake.createStatement({sqlText: sql_cmd, binds:[INPUT_TABLE]});
sql_stmt.execute();
} catch(err) {
result += "Message: " + err.message;
}
return result;
$$;
Call spExample('PAIDBILLS');
For me, this worked, first Quotes then brackets. I think its different for different type of query , other answers worked for me for select but not for SHOW
var sql_cmd = "SHOW WAREHOUSES like '(?)';"
CREATE OR REPLACE PROCEDURE spExample(INPUT_TABLE VARCHAR)
RETURNS varchar
LANGUAGE JAVASCRIPT
execute as owner
AS
$$
result = "";
try {
var sql_cmd = "SHOW WAREHOUSES like '(?)';";
var sql_stmt = snowflake.execute({sqlText: sql_cmd, binds:[INPUT_TABLE]});
} catch(err) {
result += "Message: " + err.message;
}
return result;
$$;
Call spExample('WREHOUSE NAME');
actually yes.
Bind variable is just that, a variable. So you can do a
SELECT * FRMO MY_TABLE WHERE MY_COLUMN=?
but you can't use bind to substitute for commands or column or table names. You can however use simple JS concatenation, like
var sql_cmd = "SELECT * FROM "+INPUT_TABLE;

How to debug a Snowflake stored procedure?

I am using the Snowflake Cloud Database, please help me with a tool to debugging the procedures or functions.
The following Snowflake Javascript Stored Procedure is a template I use to get started on a new Stored Procedures. It contains plenty of debugging tricks, such as:
it has a "where am I?" variable which gives you a understanding of where in the code you are
it gathers information in an array as the process moves along
it returns that array to the standard output of the call command
it has a "good start" of an exception block, who's contents also get pushed out to standard output on a call of the stored procedure, should it fail.
Something I've been meaning to add is to set a query tag in the code as well, that'd be helpful when reviewing query history, to easily identify the SQL commands that were used in the execution of the Stored Procedure.
This "ties into" the final "debugging trick" - you should always review the query history (actual queries your code executed) when developing stored procedures in a development or test environment, particularly when you are building dynamic SQL statements. Reviewing your query history is a must-do and will show you exactly the commands run and the order of operations of them running.
Here's the code with the sample table it uses, I hope it helps...Rich
CREATE OR REPLACE TABLE test_scripts (
load_seq number,
script varchar(2000)
);
INSERT INTO test_scripts values
(1, 'SELECT current_timestamp();'),
(2, 'SELECT current_warehouse();'),
(3, 'SELECT COUNT(*) FROM snowflake.account_usage.tables;'),
(4, 'SELECT current_date();'),
(5, 'SELECT current_account();'),
(6, 'SELECT COUNT(*) FROM snowflake.account_usage.tables;'),
(7, 'SELECT ''RICH'';');
select * from test_scripts;
CREATE OR REPLACE PROCEDURE sp_test(p1 varchar, p2 varchar)
RETURNS ARRAY
LANGUAGE javascript
EXECUTE AS caller
AS
$$
//note: you can change the RETURN to VARCHAR if needed
// but the array "looks nice"
try {
var whereAmI = 1;
var return_array = [];
var counter = 0;
var p1_str = "p1: " + P1
var p2_str = "p2: " + P2
var load_seq = P1;
var continue_flag = P2;
whereAmI = 2;
return_array.push(p1_str)
return_array.push(p2_str)
whereAmI = 3;
//which SQL do I want to run?
if (continue_flag=="YES") {
return_array.push("query 1")
var sqlquery = "SELECT * FROM test_scripts WHERE load_seq >= " + load_seq + " order by 1, 2;";
}
else {
return_array.push("query 2")
var sqlquery = "SELECT * FROM test_scripts WHERE load_seq = " + load_seq + " order by 1, 2;";
}
whereAmI = 4;
//begin the run of grabbing the commands
var stmt = snowflake.createStatement( {sqlText: sqlquery} );
var rs = stmt.execute();
whereAmI = 5;
// Loop through the results, processing one row at a time...
while (rs.next()) {
counter = counter + 1;
var tmp_load_seq = rs.getColumnValue(1);
var tmp_script = rs.getColumnValue(2);
var tmp_rs = snowflake.execute({sqlText: tmp_script});
tmp_rs.next();
var tmp_col1 = tmp_rs.getColumnValue(1);
return_array.push("tmp_col1: " + tmp_col1)
}
whereAmI = 6;
return_array.push("end process - counter: " + counter)
return return_array;
}
catch (err) {
return_array.push("error found")
return_array.push(whereAmI)
return_array.push(err)
return return_array;
}
$$;
CALL sp_test(3, 'NO');
I do not believe there is any editor / debugger for stored procedures for Snowflake. Few options:
You can break your code to smaller parts and try to troubleshoot
Use a log table and insert into log table often, so you can look at the log table to find out what went wrong
unfortunately there isn't one environment to rule them all
1. write your SQL in a Worksheet or Editor
2. write your SPROC code in a JS enabled editor
3. merge them together in a Worksheet or Editor
4. Unit test in SPROCS as shown above by #Rich Murmane
I normally just write SPROCS in a Worksheet but it isnt optimal
Logging is your friend here, as there is no debugger. In general finding and using a debugger for db stored procedures is hard to pull off. Not impossible, just unlikely.
This is a decent alternative:
CREATE or replace PROCEDURE do_log(MSG STRING)
RETURNS STRING
LANGUAGE JAVASCRIPT
EXECUTE AS CALLER
AS $$
//see if we should log - checks for do_log = true session variable
try{
var foo = snowflake.createStatement( { sqlText: `select $do_log` } ).execute();
} catch (ERROR){
return; //swallow the error, variable not set so don't log
}
foo.next();
if (foo.getColumnValue(1)==true){ //if the value is anything other than true, don't log
try{
snowflake.createStatement( { sqlText: `create temp table identifier ($log_table) if not exists (ts number, msg string)`} ).execute();
snowflake.createStatement( { sqlText: `insert into identifier ($log_table) values (:1, :2)`, binds:[Date.now(), MSG] } ).execute();
} catch (ERROR){
throw ERROR;
}
}
$$
;
Then in the stored procedure, you want to debug add a log function at the top:
function log(msg){
snowflake.createStatement( { sqlText: `call do_log(:1)`, binds:[msg] } ).execute();
}
Then above the call to the stored procedure:
set do_log = true; --true to enable logging, false (or undefined) to disable
set log_table = 'my_log_table'; --The name of the temp table where log messages go
Then in the actual stored procedure you need to add some logging lines:
log('this is another log message');
Then call the stored procedure as you would normally. Then select from my_log_table.
Important note: this uses a temp table, so you won't be able to read from that logging table in a different Snowflake connection. This means if you're using the Worksheet editor you need to keep all this stuff on the same sheet.
"Borrowed" from: https://community.snowflake.com/s/article/Snowflake-Stored-Procedure-Logging

Resources