I'm trying to essentially define an always finally (similar to CTL) query operation using Drools. The tree is composed of nodes (called Artifacts) that are annotated with a path id. Every split in the tree (a parent with more than one child) is represented by generating a new path id and inserting the fact SplitFrom(child, parent) to the knowledge base.
Essentially, we want to see if, from some starting path id, a given Artifact object exists on all paths in the tree. My attempt to o this is shown below:
query alwaysFinally( String $type, String $productName, long $parentPathId )
Artifact( type == $type, productName == $productName, pathId == $parentPathId )
or
forall( SplitFrom( parent == $parentPathId, $childPathId := child )
and
?alwaysFinally( $type, $productName, $childPathId; ) )
end
Unfortunately, this results in the following error:
java.lang.RuntimeException: Build Errors:
Error Messages:
Message [id=1, level=ERROR, path=edu/umn/crisys/sim/agent/cognition/futureworld/rules/process/antecedent/commonAntecedents.drl, line=47, column=0
text=[ERR 102] Line 47:6 mismatched input '?' in query]
Message [id=2, level=ERROR, path=edu/umn/crisys/sim/agent/cognition/futureworld/rules/process/antecedent/commonAntecedents.drl, line=0, column=0
text=Parser returned a null Package]
...
I've played with inserting parenthesis in a number of different ways, but I don't think that is the real problem. If I remove the and and replace it with a comma or a newline, I get the following error:
java.lang.ClassCastException: org.drools.core.rule.QueryElement cannot be cast to org.drools.core.rule.Pattern
at org.drools.compiler.rule.builder.ForallBuilder.build(ForallBuilder.java:57)
at org.drools.compiler.rule.builder.ForallBuilder.build(ForallBuilder.java:32)
at org.drools.compiler.rule.builder.GroupElementBuilder.build(GroupElementBuilder.java:66)
at org.drools.compiler.rule.builder.GroupElementBuilder.build(GroupElementBuilder.java:36)
at org.drools.compiler.rule.builder.GroupElementBuilder.build(GroupElementBuilder.java:66)
at org.drools.compiler.rule.builder.RuleBuilder.build(RuleBuilder.java:97)
at org.drools.compiler.builder.impl.KnowledgeBuilderImpl.addRule(KnowledgeBuilderImpl.java:1820)
at org.drools.compiler.builder.impl.KnowledgeBuilderImpl.compileRules(KnowledgeBuilderImpl.java:1111)
at org.drools.compiler.builder.impl.KnowledgeBuilderImpl.compileAllRules(KnowledgeBuilderImpl.java:989)
at org.drools.compiler.builder.impl.CompositeKnowledgeBuilderImpl.buildRules(CompositeKnowledgeBuilderImpl.java:260)
at org.drools.compiler.builder.impl.CompositeKnowledgeBuilderImpl.buildPackages(CompositeKnowledgeBuilderImpl.java:121)
at org.drools.compiler.builder.impl.CompositeKnowledgeBuilderImpl.build(CompositeKnowledgeBuilderImpl.java:105)
at org.drools.compiler.kie.builder.impl.AbstractKieModule.buildKnowledgePackages(AbstractKieModule.java:243)
at org.drools.compiler.kie.builder.impl.AbstractKieProject.verify(AbstractKieProject.java:64)
at org.drools.compiler.kie.builder.impl.KieBuilderImpl.buildKieProject(KieBuilderImpl.java:230)
at org.drools.compiler.kie.builder.impl.KieBuilderImpl.buildAll(KieBuilderImpl.java:198)
I don't think accumulators will work here since a query is not a Pattern object.
Does anyone have a good idea of how to express this in Drools?
You don't need the forall. This is something that Drools does automatically for all matches, to continue the matching process. forall is the universal quantifier which you don't need here since the condition need not be true for all of a set of facts.
query alwaysFinally( String $type, String $productName, long $parentPathId )
Artifact( type == $type, productName == $productName, pathId == $parentPathId )
or
( SplitFrom( parent == $parentPathId, $childPathId := child )
and
?alwaysFinally( $type, $productName, $childPathId; )
end
It may result in multiple matches if the Artifact occurs more than once.
Logic rules won out. The latter part of the "or" statement can be represented as (using set notation)
let childPathIds: { SplitFrom(parent == $parentPathId, child := childPathId }
forall x in childPathIds, P(x)
Now, we can doubly negate and preserve identity. This gets us:
let childPathIds: { SplitFrom(parent == $parentPathId, child := childPathId }
exists x in childPathIds, not P(x)
Thus, it can be represented in dools using the following
accumulate( SplitFrom( parent == $parentPathId, $childPathId := child );
$childPaths: collectSet( $childPathId ) )
and
// Set non-empty
exists( $childPathId: Long() from $childPaths )
and
( not( exists( $childPathId: Long( ) from $childPaths
and
not (?alwaysFinally( $type, $productName, $childPathId; ) ) ) )
)
Notice the additional non-empty check? We need that because the 3rd condition will return true if $childPaths is empty.
So, moral of the story: Yes, we can capture forall with a subquery; however we have to use a much more clunky syntax to do it. At least we don't have to go down to Java to implement it.
Related
I have the following Go struct:
type Bar struct {
Stuff string `db:"stuff"`
Other string `db:"other"`
}
type Foo struct {
ID int `db:"id"`
Bars []*Bar `db:"bars"`
}
So Foo contains a slice of Bar pointers. I also have the following tables in Postgres:
CREATE TABLE foo (
id INT
)
CREATE TABLE bar (
id INT,
stuff VARCHAR,
other VARCHAR,
trash VARCHAR
)
I want to LEFT JOIN on table bar and aggregate it as an array to be stored in the struct Foo. I've tried:
SELECT f.*,
ARRAY_AGG(b.stuff, b.other) AS bars
FROM foo f
LEFT JOIN bar b
ON f.id = b.id
WHERE f.id = $1
GROUP BY f.id
But it looks like the ARRAY_AGG function signature is incorrect (function array_agg(character varying, character varying) does not exist). Is there a way to do this without making a separate query to bar?
It looks like what you want is for bars to be an array of bar objects to match your Go types. To do this, you should use JSON_AGG rather than ARRAY_AGG since ARRAY_AGG only works on single columns and would produce in this case an array of type text (TEXT[]). JSON_AGG, on the other hand, creates an array of json objects. You can combine this with JSON_BUILD_OBJECT to select only the columns you want.
Here's an example:
SELECT f.*,
JSON_AGG(JSON_BUILD_OBJECT('stuff', b.stuff, 'other', b.other)) AS bars
FROM foo f
LEFT JOIN bar b
ON f.id = b.id
WHERE f.id = $1
GROUP BY f.id
Then you'll have to handle unmarshaling the json in Go, but other than that you should be good to go.
Note also that Go will ignore unused keys for you when unmarshaling json to a struct, so you can simplify the query by just selecting all fields on the bar table if you want. Like so:
SELECT f.*,
JSON_AGG(TO_JSON(b.*)) AS bars -- or JSON_AGG(b.*)
FROM foo f
LEFT JOIN bar b
ON f.id = b.id
WHERE f.id = $1
GROUP BY f.id
If you want to also handle cases where there are no entries in bar for a record in foo, you can use:
SELECT f.*,
COALESCE(
JSON_AGG(TO_JSON(b.*)) FILTER (WHERE b.id IS NOT NULL),
'[]'::JSON
) AS bars
FROM foo f
LEFT JOIN bar b
ON f.id = b.id
WHERE f.id = $1
GROUP BY f.id
Without the FILTER, you'll get [NULL] for rows in foo that have no corresponding rows in bar, and the FILTER gives you just NULL instead, then just use COALESCE to convert to an empty json array.
As you already know array_agg takes a single argument and returns an array of the type of the argument. So, if you want all of a row's columns to be included in the array's elements you can just pass in the row reference directly, e.g.:
SELECT array_agg(b) FROM b
If, however, you only want to include specific columns in the array's elements you can use the ROW constructor, e.g.:
SELECT array_agg(ROW(b.stuff, b.other)) FROM b
Go's standard library provides out-of-the-box support for scanning only scalar values. For scanning more complex values like arbitrary objects and arrays one has to either look for 3rd party solutions, or implement their own sql.Scanner.
To be able to implement your own sql.Scanner and properly parse a postgres array of rows you first need to know what format postgres uses to output the value, you can find this out by using psql and some queries directly:
-- simple values
SELECT ARRAY[ROW(123,'foo'),ROW(456,'bar')];
-- output: {"(123,foo)","(456,bar)"}
-- not so simple values
SELECT ARRAY[ROW(1,'a b'),ROW(2,'a,b'),ROW(3,'a",b'),ROW(4,'(a,b)'),ROW(5,'"','""')];
-- output: {"(1,\"a b\")","(2,\"a,b\")","(3,\"a\"\",b\")","(4,\"(a,b)\")","(5,\"\"\"\",\"\"\"\"\"\")"}
As you can see this can get pretty hairy but nevertheless it's parseable, the syntax looks to be something like this:
{"(column_value[, ...])"[, ...]}
where column_value is either an unquoted value, or a quoted value with escaped double quotes, and such a quoted value itself can contain escaped double quotes but only in twos, i.e. a single escaped double quote will not occur inside the column_value. So a rough and incomplete implementation of the parser might look something like this:
NOTE: there may be other syntax rules, that I do not know of, that need to be taken into consideration during parsing. In addition to that the code below doesn't handle NULLs properly.
func parseRowArray(a []byte) (out [][]string) {
a = a[1 : len(a)-1] // drop surrounding curlies
for i := 0; i < len(a); i++ {
if a[i] == '"' { // start of row element
row := []string{}
i += 2 // skip over current '"' and the following '('
for j := i; j < len(a); j++ {
if a[j] == '\\' && a[j+1] == '"' { // start of quoted column value
var col string // column value
j += 2 // skip over current '\' and following '"'
for k := j; k < len(a); k++ {
if a[k] == '\\' && a[k+1] == '"' { // end of quoted column, maybe
if a[k+2] == '\\' && a[k+3] == '"' { // nope, just escaped quote
col += string(a[j:k]) + `"`
k += 3 // skip over `\"\` (the k++ in the for statement will skip over the `"`)
j = k + 1 // skip over `\"\"`
continue // go to k loop
} else { // yes, end of quoted column
col += string(a[j:k])
row = append(row, col)
j = k + 2 // skip over `\"`
break // go back to j loop
}
}
}
if a[j] == ')' { // row end
out = append(out, row)
i = j + 1 // advance i to j's position and skip the potential ','
break // go to back i loop
}
} else { // assume non quoted column value
for k := j; k < len(a); k++ {
if a[k] == ',' || a[k] == ')' { // column value end
col := string(a[j:k])
row = append(row, col)
j = k // advance j to k's position
break // go back to j loop
}
}
if a[j] == ')' { // row end
out = append(out, row)
i = j + 1 // advance i to j's position and skip the potential ','
break // go to back i loop
}
}
}
}
}
return out
}
Try it on playground.
With something like that you can then implement an sql.Scanner for your Go slice of bars.
type BarList []*Bar
func (ls *BarList) Scan(src interface{}) error {
switch data := src.(type) {
case []byte:
a := praseRowArray(data)
res := make(BarList, len(a))
for i := 0; i < len(a); i++ {
bar := new(Bar)
// Here i'm assuming the parser produced a slice of at least two
// strings, if there are cases where this may not be the true you
// should add proper length checks to avoid unnecessary panics.
bar.Stuff = a[i][0]
bar.Other = a[i][1]
res[i] = bar
}
*ls = res
}
return nil
}
Now if you change the type of the Bars field in the Foo type from []*Bar to BarList you'll be able to directly pass in a pointer of the field to a (*sql.Row|*sql.Rows).Scan call:
rows.Scan(&f.Bars)
If you don't want to change the field's type you can still make it work by converting the pointer just when it's being passed to the Scan method:
rows.Scan((*BarList)(&f.Bars))
JSON
An sql.Scanner implementation for the json solution suggested by Henry Woody would look something like this:
type BarList []*Bar
func (ls *BarList) Scan(src interface{}) error {
if b, ok := src.([]byte); ok {
return json.Unmarshal(b, ls)
}
return nil
}
I have a SQL query which takes user inputs hence security flaw is present.
The existing query is:
SELECT BUS_NM, STR_ADDR_1, CITY_NM, STATE_CD, POSTAL_CD, COUNTRY_CD,
BUS_PHONE_NB,PEG_ACCOUNT_ID, GDN_ALERT_ID, GBIN, GDN_MON_REF_NB,
ALERT_DT, ALERT_TYPE, ALERT_DESC,ALERT_PRIORITY
FROM ( SELECT A.BUS_NM, AE.STR_ADDR_1, A.CITY_NM, A.STATE_CD, A.POSTAL_CD,
CC.COUNTRY_CD, A.BUS_PHONE_NB, A.PEG_ACCOUNT_ID, 'I' ||
LPAD(INTL_ALERT_DTL_ID, 9,'0') GDN_ALERT_ID,
LPAD(IA.GBIN, 9,'0') GBIN, IA.GDN_MON_REF_NB,
DATE(IAD.ALERT_TS) ALERT_DT,
XMLCAST(XMLQUERY('$A/alertTypeConfig/biqCode/text()' passing
IAC.INTL_ALERT_TYPE_CONFIG as "A") AS CHAR(4)) ALERT_TYPE,
, ROW_NUMBER() OVER () AS "RN"
FROM ACCOUNT A, Other tables
WHERE IA.GDN_MON_REF_NB = '100'
AND A.PEG_ACCOUNT_ID = IAAR.PEG_ACCOUNT_ID
AND CC.COUNTRY_CD = A.COUNTRY_ISO3_CD
ORDER BY IA.INTL_ALERT_ID ASC )
WHERE ALERT_TYPE IN (" +TriggerType+ ");
I changed it to accept TriggerType from setString like:
SELECT BUS_NM, STR_ADDR_1, CITY_NM, STATE_CD, POSTAL_CD, COUNTRY_CD,
BUS_PHONE_NB,PEG_ACCOUNT_ID, GDN_ALERT_ID, GBIN, GDN_MON_REF_NB,
ALERT_DT, ALERT_TYPE, ALERT_DESC,ALERT_PRIORITY
FROM ( SELECT A.BUS_NM, AE.STR_ADDR_1, A.CITY_NM, A.STATE_CD, A.POSTAL_CD,
CC.COUNTRY_CD, A.BUS_PHONE_NB, A.PEG_ACCOUNT_ID,
'I' || LPAD(INTL_ALERT_DTL_ID, 9,'0') GDN_ALERT_ID,
LPAD(IA.GBIN, 9,'0') GBIN, IA.GDN_MON_REF_NB,
DATE(IAD.ALERT_TS) ALERT_DT,
XMLCAST(XMLQUERY('$A/alertTypeConfig/biqCode/text()' passing
IAC.INTL_ALERT_TYPE_CONFIG as "A") AS CHAR(4)) ALERT_TYPE,
ROW_NUMBER() OVER () AS "RN"
FROM ACCOUNT A, other tables
WHERE IA.GDN_MON_REF_NB = '100'
AND A.PEG_ACCOUNT_ID = IAAR.PEG_ACCOUNT_ID
AND CC.COUNTRY_CD = A.COUNTRY_ISO3_CD
ORDER BY IA.INTL_ALERT_ID ASC )
WHERE ALERT_TYPE IN (?);
Setting trigger type as below:
if (StringUtils.isNotBlank(request.getTriggerType())) {
preparedStatement.setString(1, triggerType != null ? triggerType.toString() : "");
}
Getting error as
Caused by: com.ibm.db2.jcc.am.SqlDataException: DB2 SQL Error: SQLCODE=-302, SQLSTATE=22001, SQLERRMC=null, DRIVER=4.19.26
The -302 SQLCODE indicates a conversion error of some sort.
SQLSTATE 22001 narrows that down a bit by telling us that you are trying to force a big string into a small variable. Given the limited information in your question, I am guessing it is the XMLCAST that is the culprit.
DB2 won't jam 30 pounds of crap into a 4 pound bag so to speak, it gives you an error. Maybe giving XML some extra room in the cast might be a help. If you need to make sure it ends up being only 4 characters long, you could explicitly do a LEFT(XMLCAST( ... AS VARCHAR(64)), 4). That way the XMLCAST has the space it needs, but you cut it back to fit your variable on the fetch.
The other thing could be that the variable being passed to the parameter marker is too long. DB2 will guess the type and length based on the length of ALERT_TYPE. Note that you can only pass a single value through a parameter marker. If you pass a comma separated list, it will not behave as expected (unless you expect ALERT_TYPE to also contain a comma separated list). If you are getting the comma separated list from a table, you can use a sub-select instead.
Wrong IN predicate use with a parameter.
Do not expect that IN ('AAAA, M250, ABCD') (as you try to do passing a comma-separated string as a single parameter) works as IN ('AAAA', 'M250', 'ABCD') (as you need). These predicates are not equivalent.
You need some "string tokenizer", if you want to pass such a comma-separated string like below.
select t.*
from
(
select XMLCAST(XMLQUERY('$A/alertTypeConfig/biqCode/text()' passing IAC.INTL_ALERT_TYPE_CONFIG as "A") AS CHAR(4)) ALERT_TYPE
from table(values xmlparse(document '<alertTypeConfig><biqCode>M250, really big code</biqCode></alertTypeConfig>')) IAC(INTL_ALERT_TYPE_CONFIG)
) t
--WHERE ALERT_TYPE IN ('AAAA, M250, ABCD')
join xmltable('for $id in tokenize($s, ",\s?") return <i>{string($id)}</i>'
passing cast('AAA, M250 , ABCD' as varchar(200)) as "s"
columns token varchar(200) path '.') x on x.token=t.ALERT_TYPE
;
Run the statement as is. Then you may uncomment the string with WHERE clause and comment out the rest to see what you try to do.
P.S.:
The error you get is probably because you don't specify the data type of the parameter (you don't use something like IN (cast(? as varchar(xxx))), and db2 compiler assumes that its length is equal to the length of the ALERT_TYPE expression (4 bytes).
I have the following select statement in ABAP:
SELECT munic~mandt VREFER BIS AB ZZELECDATE ZZCERTDATE CONSYEAR ZDIMO ZZONE_M ZZONE_T USAGE_M USAGE_T M2MC M2MT M2RET EXEMPTMCMT EXEMPRET CHARGEMCMT
INTO corresponding fields of table GT_INSTMUNIC_F
FROM ZCI00_INSTMUNIC AS MUNIC
INNER JOIN EVER AS EV on
MUNIC~POD = EV~VREFER(9).
"where EV~BSTATUS = '14' or EV~BSTATUS = '32'.
My problem with the above statement is that does not recognize the substring/offset operation on the 'ON' clause. If i remove the '(9) then
it recognizes the field, otherwise it gives error:
Field ev~refer is unknown. It is neither in one of the specified tables
nor defined by a "DATA" statement. I have also tried doing something similar in the 'Where' clause, receiving a similar error:
LOOP AT gt_instmunic.
clear wa_gt_instmunic_f.
wa_gt_instmunic_f-mandt = gt_instmunic-mandt.
wa_gt_instmunic_f-bis = gt_instmunic-bis.
wa_gt_instmunic_f-ab = gt_instmunic-ab.
wa_gt_instmunic_f-zzelecdate = gt_instmunic-zzelecdate.
wa_gt_instmunic_f-ZZCERTDATE = gt_instmunic-ZZCERTDATE.
wa_gt_instmunic_f-CONSYEAR = gt_instmunic-CONSYEAR.
wa_gt_instmunic_f-ZDIMO = gt_instmunic-ZDIMO.
wa_gt_instmunic_f-ZZONE_M = gt_instmunic-ZZONE_M.
wa_gt_instmunic_f-ZZONE_T = gt_instmunic-ZZONE_T.
wa_gt_instmunic_f-USAGE_M = gt_instmunic-USAGE_M.
wa_gt_instmunic_f-USAGE_T = gt_instmunic-USAGE_T.
temp_pod = gt_instmunic-pod.
SELECT vrefer
FROM ever
INTO wa_gt_instmunic_f-vrefer
WHERE ( vrefer(9) LIKE temp_pod ). " PROBLEM WITH SUBSTRING
"AND ( BSTATUS = '14' OR BSTATUS = '32' ).
ENDSELECT.
WRITE: / sy-dbcnt.
WRITE: / 'wa is: ', wa_gt_instmunic_f.
WRITE: / 'wa-ever is: ', wa_gt_instmunic_f-vrefer.
APPEND wa_gt_instmunic_f TO gt_instmunic_f.
WRITE: / wa_gt_instmunic_f-vrefer.
ENDLOOP.
itab_size = lines( gt_instmunic_f ).
WRITE: / 'Internal table populated with', itab_size, ' lines'.
The basic task i want to implement is to modify a specific field on one table,
pulling values from another. They have a common field ( pod = vrefer(9) ). Thanks in advance for your time.
If you are on a late enough NetWeaver version, it works on 7.51, you can use the OpenSQL function LEFT or SUBSTRING. Your query would look something like:
SELECT munic~mandt VREFER BIS AB ZZELECDATE ZZCERTDATE CONSYEAR ZDIMO ZZONE_M ZZONE_T USAGE_M USAGE_T M2MC M2MT M2RET EXEMPTMCMT EXEMPRET CHARGEMCMT
FROM ZCI00_INSTMUNIC AS MUNIC
INNER JOIN ever AS ev
ON MUNIC~POD EQ LEFT( EV~VREFER, 9 )
INTO corresponding fields of table GT_INSTMUNIC_F.
Note that the INTO clause needs to move to the end of the command as well.
field(9) is a subset operation that is processed by the ABAP environment and can not be translated into a database-level SQL statement (at least not at the moment, but I'd be surprised if it ever will be). Your best bet is either to select the datasets separately and merge them manually (if both are approximately equally large) or pre-select one and use a FAE/IN clause.
They have a common field ( pod = vrefer(9) )
This is a wrong assumption, because they both are not fields, but a field an other thing.
If you really need to do that task through SQL, I'll suggest you to check native SQL sentences like SUBSTRING and check if you can manage to use them within an EXEC_SQL or (better) the CL_SQL* classes.
I'm working on a report in Crystal Reports XI that allows someone to filter help desk tickets using a number of optional dynamic parameters. If I make a selection for each parameter it returns the expected results, but if I leave out any of the parameters it doesn't return anything and when I look at the SQL query it says, "No SQL Query is used because the record selection formula returns no records." I currently have the following code for Record Selection:
{Incident.State:} = "C" and
{Incident.Close Date & Time} in {?BDate} to {?EDate} and
If HasValue({?Group}) Then (
{Groups.Code} = {?Group}
)
and
If HasValue({?Category}) Then (
{Incident.Subject Description} = {?Category}
)
and
If HasValue({?Staff}) Then (
{Incident_Details.Login ID} = {?Staff}
)
and
If HasValue({?Community}) Then (
{Incident.Company Name} = {?Community}
)
To me, this seems like it should work and if I leave out the If statement to verify the parameters have values I get an error so it seems like the HasValue is working properly. Any ideas?
Your problem stems from not explicitly handling the optional parameter configurations. It's easy to run into this problem when using if-statements in the record selection formula. Instead, explicitly handle the cases where the parameters DO and DO NOT have values. Something like this:
...
(not(hasvalue({?Group})) or {Groups.Code}={?Group})
and
(not(hasvalue({?Category})) or {Incident.Subject Description} = {?Category})
and
(not(hasvalue({?Staff})) or {Incident_Details.Login ID} = {?Staff} )
and
(not(hasvalue({?Community})) or {Incident.Company Name} = {?Community})
Doing it this way effectively tells CR to just ignore the parameter if is doesn't have a value, otherwise select records based on what was entered in those parameters.
The Record Selection formula indicates whether a record must be used, or not, by returning true or false. With that in mind, if some value is informed, the formula must return True if it meets some criteria. If nothing is informed on the filter, the formula must always return True.
Try something like this:
{Incident.State:} = "C" and
{Incident.Close Date & Time} in {?BDate} to {?EDate} and
(If HasValue({?Group}) Then (
{Groups.Code} = {?Group}
) else
(True))
and
(If HasValue({?Category}) Then (
{Incident.Subject Description} = {?Category}
) else
(True))
and (
If HasValue({?Staff}) Then (
{Incident_Details.Login ID} = {?Staff}
) else
(True))
and (
If HasValue({?Community}) Then (
{Incident.Company Name} = {?Community}
) else
(True))
I remember having problem with DBI method selectrow_array. When i wasn't tidy enough i got back from it not the value of the column i asked, but count of columns (or something unwanted, i can't recall exactly). Now i try to refactor some code and i want to make sure in every possible place, that i get back only expected value. So i try to avoid surprises and find out which the bad behaviour was. From DBI docs i read that this may be really be problematic situation:
If called in a scalar context for a statement handle that has more
than one column, it is undefined whether the driver will return the
value of the first column or the last. So don't do that. Also, in a
scalar context, an "undef" is returned if there are no more rows or if
an error occurred. That "undef" can't be distinguished from an "undef"
returned because the first field value was NULL. For these reasons
you should exercise some caution if you use "selectrow_array" in a
scalar context, or just don't do that.
Still i can't force selectrow_array to return anything but value of the col1 (that's it what i am expecting)
my $query = 'SELECT col1, col2, col3 FROM table WHERE id = 112233';
my ( $c ) = ( $dbh->selectrow_array( $query ) );
my $x = ask_from_db();
my $y = $dbh->selectrow_array( $query );
my $z = ( $dbh->selectrow_array( $query ) );
my #A = $dbh->selectrow_array( $query );
say "C: $c"; # C: col1
say "X: $x"; # X: col1
say "Y: $y"; # Y: col1
say "Z: $z"; # Z: col1
say "A: #A"; # A: col1 col2 col3
sub ask_from_db {
return $dbh->selectrow_array( $query );
}
Every way i ask above, gives me fine result. How should i run the query to get wrong result?
wrong result != col1 value
The difference in outcome will be based on the implementation of the driver.
wantarray ? #row : $row[0]
vs
wantarray ? #row : $row[-1]
You'd use to use a different driver to get a different outcome. That said, I imagine you'll have a hard time finding a driver that doesn't return the first.
If you want to be sure to get the first, use:
( $dbh->selectrow_array( $query ) )[0]
What the documentation means by "it is undefined whether the driver will return the value of the first column or the last" is that the column returned is defined by the database driver and not DBI.
So the Postgres driver may decide to always return the first column whereas the mysql driver may always return the last column, or the column returned might depend on the query.
So don't call selectrow_array is scalar context - always call it in list context:
my #row = $sth->selectrow_array($query)
and you'll avoid all of the issues that the documentation mentions.