I would like to query an SQLite table that contains directory paths to find all the paths under some hierarchy. Here's an example of the contents of the column:
/alpha/papa/
/alpha/papa/tango/
/alpha/quebec/
/bravo/papa/
/bravo/papa/uniform/
/charlie/quebec/tango/
If I search for everything under /bravo/papa/, I would like to get:
/bravo/papa/
/bravo/papa/uniform/
I am currently trying to do this like so (see below for the long story of why I can't use more simple methods):
SELECT * FROM Files WHERE Path >= '/bravo/papa/' AND Path < '/bravo/papa0';
This works. It looks a bit weird, but it works for this example. '0' is the unicode code point 1 greater than '/'. When ordered lexicographically, all the paths starting with '/bravo/papa/' compare greater than it and less than 'bravo/papa0'. However, in my tests, I find that this breaks down when we try this:
SELECT * FROM Files WHERE Path >= '/' AND Path < '0';
This returns no results, but it should return every row. As far as I can tell, the problem is that SQLite is treating '0' as a number, not a string. If I use '0Z' instead of '0', for example, I do get results, but I introduce a risk of getting false positives. (For example, if there actually was an entry '0'.)
The simple version of my question is: is there some way to get SQLite to treat '0' in such a query as the length-1 string containing the unicode character '0' (which should sort strings such as '!', '*' and '/', but before '1', '=' and 'A') instead of the integer 0 (which SQLite sorts before all strings)?
I think in this case I can actually get away with special-casing a search for everything under '/', since all my entries will always start with '/', but I'd really like to know how to avoid this sort of thing in general, as it's unpleasantly surprising in all the same ways as Javascript's "==" operator.
First approach
A more natural approach would be to use the LIKE or GLOB operator. For example:
SELECT * FROM Files WHERE Path LIKE #prefix || '%';
But I want to support all valid path characters, so I would need to use ESCAPE for the '_' and '%' symbols. Apparently this prevents SQLite from using an index on Path. (See http://www.sqlite.org/optoverview.html#like_opt ) I really want to be able to benefit from an index here, and it sounds like that's impossible using either LIKE or GLOB unless I can guarantee that none of their special characters will occur in the directory name, and POSIX allows anything other than NUL and '/', even GLOB's '*' and '?' characters.
I'm providing this for context. I'm interested in other approaches to solve the underlying problem, but I'd prefer to accept an answer that directly addresses the ambiguity of strings-that-look-like-numbers in SQLite.
Similar questions
How do I prevent sqlite from evaluating a string as a math expression?
In that question, the values weren't quoted. I get these results even when the values are quoted or passed in as parameters.
EDIT - See my answer below. The column was created with the invalid type "STRING", which SQLite treated as NUMERIC.
* Groan *. The column had NUMERIC affinity because it had accidentally been specified as "STRING" instead of "TEXT". Since SQLite didn't recognize the type name, it made it NUMERIC, and because SQLite doesn't enforce column types, everything else worked as expected, except that any time a number-like string is inserted into that column it is converted into a numeric type.
Related
In SQLite you can use named parameters in statements, like this (Python example):
cur.execute("insert into lang values (:foo, :bar)", {'foo': 'a', 'bar': 2})
Is there any way to have parameter names containing spaces? I.e:
cur.execute("insert into lang values (:'foo bar')", {'foo bar': 'a'})
The documentation suggests not but you never know.
Apparently for the #AAA form you can:
The identifier name in this case can include one or more occurrences of "::" and a suffix enclosed in "(...)" containing any text at all.
But that doesn't let you have an arbitrary name since the brackets are still part of the name. So the answer appears to be no.
In my database, a table contains two columns each containing an 8 digit ASCII code, usually it's just alphanumeric. For example, a row might contain A123B45C in col1 and PQ2R4680 in col2.
I need to have a query/view that outputs a 4 character string calculated as the 2nd+3rd chars of these, concatenated. So in this example the extra column value would be 12Q2.
This is a cut-down version of the SQL I'd like to use, although it won't work as written because of zero stripping / conversion:
select
*,
(substr(col1, 2, 2) || substr(col2, 2, 2)) AS mode
from (nested SQL source query)
where (conditions)
This fails because if a row contains A00B23B4 in col1 and P32R4680 in col2, it will evaluate as 0032 and the query output will contain numeric 32 not 0032. (It's worse if col1 contains P1-2345 or "1.23456" or something like that)
Other questions on preventing zero stripping and string to integer conversion in Sqlite, all relate to data in tables where you can define a column text affinity, or static (quotable) data. In this case I can't do these things. I also can only create queries, not tables, so I can't write to a temp table.
What is the best way to ensure I get a 4 character output in all cases?
I believe you issue is not with substr stripping characters as this works as expected e.g. :-
Then running query SELECT substr(col1,2,2) || substr(col2,2,2) as mode FROM stripping
results in (as expected):-
Rather, your issue is likely how you subsequently utilise mode in which case you may need to use a CAST expression CAST expressions
For example the following does what is possibly happening :-
`SELECT substr(col1,2,2) || substr(col2,2,2) as mode, CAST(substr(col1,2,2) || substr(col2,2,2) AS INTEGER) AS oops FROM stripping`
resulting in :-
How performant is the SQLite3 REGEXP operator?
For simplicity, assume a simple table with a single column pattern and an index
CREATE TABLE `foobar` (`pattern` TEXT);
CREATE UNIQUE INDEX `foobar_index` ON `foobar`(`pattern`);
and a query like
SELECT * FROM `foobar` WHERE `pattern` REGEXP 'foo.*'
I have been trying to compare and understand the output from EXPLAIN and it seems to be similar to using LIKE except it will be using regexp for matching. However, I am not fully sure how to read the output from EXPLAIN and I'm not getting a grasp of how performant it will be.
I understand it will be slow compared to a indexed WHERE `pattern` = 'foo' query but is it slower/similar to LIKE?
sqlite does not optimize WHERE ... REGEXP ... to use indexes. x REGEXP y is simply a function call; it's equivalent to regexp(x,y). Also note that not all installations of sqlite have a regexp function defined so using it (or the REGEXP operator) is not very portable. LIKE/GLOB on the other hand can take advantage of indexes for prefix queries provided that some additional conditions are met:
The right-hand side of the LIKE or GLOB must be either a string literal or a parameter bound to a string literal that does not begin with a wildcard character.
It must not be possible to make the LIKE or GLOB operator true by having a numeric value (instead of a string or blob) on the left-hand side. This means that either:
the left-hand side of the LIKE or GLOB operator is the name of an indexed column with TEXT affinity, or
the right-hand side pattern argument does not begin with a minus sign ("-") or a digit.
This constraint arises from the fact that numbers do not sort in lexicographical order. For example: 9<10 but '9'>'10'.
The built-in functions used to implement LIKE and GLOB must not have been overloaded using the sqlite3_create_function() API.
For the GLOB operator, the column must be indexed using the built-in BINARY collating sequence.
For the LIKE operator, if case_sensitive_like mode is enabled then the column must indexed using BINARY collating sequence, or if case_sensitive_like mode is disabled then the column must indexed using built-in NOCASE collating sequence.
If the ESCAPE option is used, the ESCAPE character must be ASCII, or a single-byte character in UTF-8.
Is there a way to modify the elements a sequence so only collated versions of the items are returned?
let $currencies := ('dollar', 'Dollar', 'dollar ')
return fn:collated-only($currencies, "http://marklogic.com/collation/en/S1/T00BB/AS")
=> ('dollar', 'dollar', 'dollar')
The values that are stored in the range index (that feeds the facets) are literally the first value that was encountered that compared equal to the others. (Because, the collation says you don't care...)
You can get a long way by calling
fn:replace(fn:lower-case(xdmp:diacritic-less(fn:normalize-unicode($str,"NFKC"))),"\p{P}","")
This won't be exactly the same in that it overfolds some things and underfolds others, but it may be good for your purposes.
Is this the expected output? There is no fn:collated-only function, so I'm assuming you're asking how to write such a function or whether there is such a function.
The thing is, there isn't a mapping from one string to another in collation comparisons, there is only a comparison algorithm (the Unicode Collation Algorithm) so there really is no canonical kind of string to return to you, and therefore no API to do so.
Stepping back, what is the problem you are actually trying to solve? By the rules of that collation, "dollar" and "Dollar" are equivalent, and by using it you declare you don't care which form you use, so you could use either one.
If these values are in XML elements and you have a range index using http://marklogic.com/collation/en/S1/T00BB/AS, you can do something like this:
let $ref := cts:element-reference(xs:QName("currency"), "collation=http://marklogic.com/collation/en/S1/T00BB/AS")
for $curr in cts:values($ref, (), "frequency-order")
return $curr || ": " || cts:frequency($curr)
This will produce results like:
"dollar: 15",
"euro: 12"
... and so on. The collation will disregard the differences among your sample inputs. These results could be formatted however you want. Is that what you're looking to do?
I am trying to write a custom report in Spiceworks, which uses SQLite queries. This report will fetch me hard drive serial numbers that are unfortunately stored in a few different ways depending on what version of Windows and WMI were on the machine.
Three common examples (which are enough to get to the actual question) are as follows:
Actual serial number: 5VG95AZF
Hexadecimal string with leading spaces: 2020202057202d44585730354341543934383433
Hexadecimal string with leading zeroes: 3030303030303030313131343330423137454342
The two hex strings are further complicated in that even after they are converted to ASCII representation, each pair of numbers are actually backwards. Here is an example:
3030303030303030313131343330423137454342 evaluates to 00000000111430B17ECB
However, the actual serial number on that hard drive is 1141031BE7BC, without leading zeroes and with the bytes swapped around. According to other questions and answers I have read on this site, this has to do with the "endianness" of the data.
My temporary query so far looks something like this (shortened to only the pertinent section):
SELECT pd.model as HDModel,
CASE
WHEN pd.serial like "30303030%" THEN
cast(('X''' || pd.serial || '''') as TEXT)
WHEN pd.serial like "202020%" THEN
LTRIM(X'2020202057202d44585730354341543934383433')
ELSE
pd.serial
END as HDSerial
The result of that query is something like this:
HDModel HDSerial
----------------- -------------------------------------------
Normal Serial 5VG95AZF
202020% test case W -DXW05CAT94843
303030% test case X'3030303030303030313131343330423137454342'
This shows that the X'....' notation style does convert into the correct (but backwards) result of W -DXW05CAT94843 when given a fully literal number (the 202020% line). However, I need to find a way to do the same thing to the actual data in the column, pd.serial, and I can't find a way.
My initial thought was that if I could build a string representation of the X'...' notation, then perhaps cast() would evaluate it. But as you can see, that just ends up spitting out X'3030303030303030313131343330423137454342' instead of the expected 00000000111430B17ECB. This means the concatenation is working correctly, but I can't find a way to evaluate it as hex the same was as in the manual test case.
I have been googling all morning to see if there is just some syntax I am missing, but the closest I have come is this concatenation using the || operator.
EDIT: Ultimately I just want to be able to have a simple case statement in my query like this:
SELECT pd.model as HDModel,
CASE
WHEN pd.serial like "30303030%" THEN
LTRIM(X'pd.serial')
WHEN pd.serial like "202020%" THEN
LTRIM(X'pd.serial')
ELSE
pd.serial
END as HDSerial
But because pd.serial gets wrapped in single quotes, it is taken as a literal string instead of taken as the data contained in that column. My hope was/is that there is just a character or operator I need to specify, like X'$pd.serial' or something.
END EDIT
If I can get past this first hurdle, my next task will be to try and remove the leading zeroes (the way LTRIM eats the leading spaces) and reverse the bytes, but to be honest, I would be content even if that part isn't possible because it wouldn't be hard to post-process this report in Excel to do that.
If anyone can point me in the right direction I would greatly appreciate it! It would obviously be much easier if I was using PHP or something else to do this processing, but because I am trying to have it be an embedded report in Spiceworks, I have to do this all in a single SQLite query.
X'...' is the binary representation in sqlite. If the values are string, you can just use them as such.
This should be a start:
sqlite> select X'3030303030303030313131343330423137454342';
00000000111430B17ECB
sqlite> select ltrim(X'3030303030303030313131343330423137454342','0');
111430B17ECB
I hope this puts you on the right path.