rocuinneagain
Forum Replies Created
-
**Untested** but a basic comsumer for testing for you would look something like:
\l ../kfk.q
kfk_cfg:(!) . flip(
(`metadata.broker.list;`localhost:9092);
(`group.id;`0);
(`fetch.wait.max.ms;`10);
(`statistics.interval.ms;`10000)
);
client:.kfk.Consumer[kfk_cfg];
topic1:`trades;
generalHelper:{[t;d]![t;();0b;key[d]!{(x;y)}'[valued;keyd]]}
castRules:`time`sym`size`side!("P"$;`$;`long$;first)
trades:([] time:`timestamp$();sym:`$();price:`float$();size:`long$();side:`char$())
topcb1:{[msg]
`trades insert generalHelper[enlist .j.k "c"$msg[`data];castRules]
}
.kfk.Metadata[client];
.kfk.Subscribe[client;enlist topic1;enlist .kfk.PARTITION_UA;enlist topcb1] -
If you review this blog post you will find it useful:
Kdb+/q Insights: Parsing JSON Files | KX
//Create a sample msg for testing
q)msg:`time`sym`price`size`side!("2024-08-25T10:50:10.743928";"ABC";117.4;67f;enlist "B")
q)msg
time | "2024-08-25T10:50:10.743928"
sym | "ABC"
price| 117.4
size | 67f
side | ,"B"
//Enlisting a dictionary beomcesa single row table
q)enlist msg
q)time sym price size side
--------------------------------------------------
"2024-08-25T10:50:10.743928" "ABC" 117.4 67 ,"B"
//The datatypes are not what we want
q)meta enlist msg
c | t f a
-----| -----
time | C
sym | C
price| f
size | f
side | C
//Use the function from the linked blog
q)generalHelper:{[t;d]![t;();0b;key[d]!{(x;y)}'[valued;keyd]]}
//Define the casting rules - adjust to what you would like
q)castRules:`time`sym`size`side!("P"$;`$;`long$;first)
//Now the msg can be handled to the datatypes we require
q)generalHelper[enlist msg;castRules]
time sym price size side
-------------------------------------------------
2024.08.25D10:50:10.743928000 ABC 117.4 67 B
q)meta generalHelper[enlist msg;castRules]
c | t f a
-----| -----
time | p
sym | s
price| f
size | j
side | c
//Create an empty table to test saving the data
q)trades:([] time:`timestamp$();sym:`$();price:`float$();size:`long$();side:`char$())
q)trades
time sym price size side
------------------------
//We can insert the rows as they arrive
q)`trades insert generalHelper[enlist msg;castRules]
,0
q)trades
time sym price size side
-------------------------------------------------
2024.08.25D10:50:10.743928000 ABC 117.4 67 B -
You could start Jupyter with a core restriction:
taskset -c 0-15 jupyter notebook
Or see this answer I posted previously: (you could choose 0-15 rather than 0,1)
https://stackoverflow.com/a/77481466/4256419
Edit the file https://github.com/KxSystems/jupyterq/blob/master/kernelspec/kernel.json
{ "argv": [ "taskset", "-c", "0,1", "q", "jupyterq_kernel.q", "-cds", "{connection_file}" ], "display_name": "Q (kdb+)", "language": "q", "env": {"JUPYTERQ_SERVERARGS":"","MPLBACKEND":"Agg"} }If you have already installed you can ensure Jupyter picks up the changes by reinstalling the kernel:
python -m jupyter kernelspec install --user --name=qpk /path/to/kernelspecAnd you can validate in a notebook with
system"taskset -pc ",string .z.i -
You can see an example of streaming data from http://www.coinapi.io over Websockets here:
For REST APIs the best library to use is kurl:
-
Install instructions: https://code.kx.com/q/learn/install/
Licencing specifics: https://code.kx.com/q/learn/licensing/
Your folder structure should look similar to:
$ tree ~/q
/home/user/q
├── m64
│ └── q
├── kc.lic
└── q.kAnd you’d set:
export QHOME="/home/user/q"
-
The error is:
‘2024.08.12T08:04:56.974 couldn’t connect to license daemon — exiting
This means the licence you have is an online-license which needs to contact a licencing server on initialisation to verify it is valid.
The network is either not available or there are firewall rules preventing this verification to be made.
More details here: https://code.kx.com/q/learn/licensing/#licensing-server-for-kdb-on-demand
-
You can use a timer and wrap the rest of the code in a function.
h:(); .z.po:{h,:x};
{system "q ",x," -p 0W &"} each ("feed1.q";"feed2.q");
.z.ts:{if[2=count h;system"t 0";main[]]}
\t 1000
main:{[] show "Rest of code" }Or move the code out to a different file and only load that once the handles are open:
h:(); .z.po:{h,:x};
{system "q ",x," -p 0W &"} each ("feed1.q";"feed2.q");
.z.ts:{if[2=count h;system"t 0";system"l main.q"]}
\t 1000 -
t:([] payment_type:10#
CARDCASH;vendor:10#abcd`e;fare:10?50;tip:10?50)
t
payment_type vendor fare tip
----------------------------
CARD a 30 39
CASH b 17 0
CARD c 23 49
CASH d 44 46
CARD e 12 41
CASH a 2 18
CARD b 36 49
CASH c 37 40
CARD d 44 24
CASH e 28 13
select from t where ({exec (tip=max tip) and (fare>avg fare) from x};([] tip;fare)) fby vendor
payment_type vendor fare tip
----------------------------
CARD a 30 39
CARD b 36 49You can pass in sub tables to aggregate on more than one column
fby – Reference – kdb+ and q documentation – kdb+ and q documentation (kx.com)
-
Sample table:
q)t:([] stock:10000?`1;AskOrder:10000?100;BidOrder:10000?100)
q)t
stock AskOrder BidOrder
-----------------------
n 85 43
p 40 15
n 9 47
j 10 64
i 87 88
i 79 87
m 46 41
h 92 63
g 86 39
......Function which returns a dictionary of percentiles:
q)f:{(`$x,/:string 1+til y)!az -1+(where deltas y xrank az:asc z),count z}Running the function returns a dictionary:
q)r:exec f["Ask_";4;AskOrder],f["Bid_";4;BidOrder] by stock from t
q)r
| AskOrder BidOrder
-| -------------------------------------------------------------------------
a| `Ask_1`Ask_2`Ask_3`Ask_4!25 50 74 99 `Bid_1`Bid_2`Bid_3`Bid_4!23 51 74 99
b| `Ask_1`Ask_2`Ask_3`Ask_4!26 49 77 99 `Bid_1`Bid_2`Bid_3`Bid_4!23 47 74 99
c| `Ask_1`Ask_2`Ask_3`Ask_4!23 49 74 99 `Bid_1`Bid_2`Bid_3`Bid_4!26 49 75 99
d| `Ask_1`Ask_2`Ask_3`Ask_4!29 53 78 99 `Bid_1`Bid_2`Bid_3`Bid_4!27 50 74 99
e| `Ask_1`Ask_2`Ask_3`Ask_4!22 46 73 99 `Bid_1`Bid_2`Bid_3`Bid_4!26 54 79 99
f| `Ask_1`Ask_2`Ask_3`Ask_4!25 49 75 99 `Bid_1`Bid_2`Bid_3`Bid_4!23 48 74 99
g| `Ask_1`Ask_2`Ask_3`Ask_4!26 49 73 99 `Bid_1`Bid_2`Bid_3`Bid_4!27 50 76 99
h| `Ask_1`Ask_2`Ask_3`Ask_4!24 51 77 99 `Bid_1`Bid_2`Bid_3`Bid_4!27 52 76 99
i| `Ask_1`Ask_2`Ask_3`Ask_4!24 48 74 99 `Bid_1`Bid_2`Bid_3`Bid_4!23 49 72 99
j| `Ask_1`Ask_2`Ask_3`Ask_4!23 46 72 99 `Bid_1`Bid_2`Bid_3`Bid_4!31 53 75 99
k| `Ask_1`Ask_2`Ask_3`Ask_4!21 44 72 99 `Bid_1`Bid_2`Bid_3`Bid_4!22 50 75 99
l| `Ask_1`Ask_2`Ask_3`Ask_4!24 49 73 99 `Bid_1`Bid_2`Bid_3`Bid_4!27 50 73 99
m| `Ask_1`Ask_2`Ask_3`Ask_4!23 45 70 99 `Bid_1`Bid_2`Bid_3`Bid_4!25 52 78 99
n| `Ask_1`Ask_2`Ask_3`Ask_4!26 52 77 99 `Bid_1`Bid_2`Bid_3`Bid_4!25 53 76 99
o| `Ask_1`Ask_2`Ask_3`Ask_4!22 50 75 99 `Bid_1`Bid_2`Bid_3`Bid_4!24 47 74 99
p| `Ask_1`Ask_2`Ask_3`Ask_4!24 51 76 99 `Bid_1`Bid_2`Bid_3`Bid_4!23 48 74 99Transforming the dictionary in to a table:
q){`stock xcols update stock:key x from (value x)[`AskOrder],'(value x)[`BidOrder]} r
stock Ask_1 Ask_2 Ask_3 Ask_4 Bid_1 Bid_2 Bid_3 Bid_4
-----------------------------------------------------
a 25 50 74 99 23 51 74 99
b 26 49 77 99 23 47 74 99
c 23 49 74 99 26 49 75 99
d 29 53 78 99 27 50 74 99
e 22 46 73 99 26 54 79 99
f 25 49 75 99 23 48 74 99
g 26 49 73 99 27 50 76 99
h 24 51 77 99 27 52 76 99
i 24 48 74 99 23 49 72 99
j 23 46 72 99 31 53 75 99
k 21 44 72 99 22 50 75 99
l 24 49 73 99 27 50 73 99
m 23 45 70 99 25 52 78 99
n 26 52 77 99 25 53 76 99
o 22 50 75 99 24 47 74 99
p 24 51 76 99 23 48 74 99 -
It is advised to avoid using name of column or globals as function param names as much as possible. Here you are seeing issues as a result of this.
In a HDB the issues caused by this scoping confusion can differ from in memory tables.
This is due to map-reduce in the HDB, which means the query execution path is more complicated.
1. type error:
In the query path date would be expected to be a list but if the function local date atom is used a type error is being raised
q){2024 select cnt:count i from Orders where date=2024.07.15}[.z.d] //Errorq){[] select cnt:count i from Orders where date=2024.07.15}[.z.d] //Okay2. count = 0:
You enlist the data param in one of the uses which means = returns a list and does not get the type error of query 1. However this list value does not make sense so an unexpected value is returned.
Returns the count of the table in the first partition of the HDB
q){2024 select cnt:count i from Orders where date=enlist 2024.07.15}[2024.07.15]Same result
q){2024 select cnt:count i from Orders where date=2024.07.15}[enlist 2024.07.15]Returns the counts of the first 2 partitions
q){2024 select cnt:{enlist count x}i from Orders where date=.z.d}[2#.z.d]You don’t even need to put in dates as the date param is being compared to itself you can put anything in:
q){2024 select cnt:{enlist count x}i from Orders where date=1b}[11b]3. Works: Not using date as a param name – no issues
-
PyKX also will convert a hsym directly to a Path type object
q).pykx.print[.pykx.eval["lambda x:x"]`:/path/to/file]
/path/to/file
q).pykx.print[.pykx.eval["lambda x:type(x)"]`:/path/to/file]
<class 'pathlib.PosixPath'>If you want a string of the Path you can call:
streamer:.pylibs.security[`:smStreamer] .pykx.eval["lambda x:str(x)"] file
-
See: Upgrading from embedPy – PyKX
PyKX converts Python strings to q symbols which differs from embedPy
q).pykx.eval["'hello'"]`
`helloPass your data back as Bytes for it to become q character vectors
q)string_to_bytes:.pykx.eval["lambda x: bytes(x, 'utf-8')"]
q)string_to_bytes[.pykx.eval["'hello'"]]`
"hello" -
In my answer I use 0N which is the long null. If your table column type for AskOrder is float then the resulting column will be of mixed type. Check by calling meta on the result. save cannot write mixed type columns like this.
A good solution is to update f to ensure it uses a null of the correct datatype:
f:{i:az -1+(where deltas y xrank az:asc z),count z;(`$x,/:string 1+til y)!i,(y-count i)#z count z}The added bit is z count z in place of 0N
Indexing out of bounds on a q vector returns nulls of the dataype of the vector:
q)(0 1)2 / Long null returned
0N
q)(0 1f)2 / Float null returned
0n -
Best would be to update the function to pad out nulls when needed:
q)f:{i:az -1+(where deltas y xrank az:asc z),count z;(`$x,/:string 1+til y)!i,(y-count i)#0N}
q)r:exec f["Ask_";6;AskOrder],f["Bid_";6;BidOrder] by stock from st
q){`stock xcols update stock:key x from (value x)[`AskOrder],'(value x)[`BidOrder]} r
stock Ask_1 Ask_2 Ask_3 Ask_4 Ask_5 Ask_6 Bid_1 Bid_2 Bid_3 Bid_4 Bid_5 Bid_6
-----------------------------------------------------------------------------
a 17 28 59 61 64 34 49 52 90 92
b 41 46 74 92 94 97 26 30 70 82 88 93
c 6 35 35 37 96 2 30 30 38 58
d 4 4 7 10 66 96 12 23 38 39 45 61
e 4 21 49 68 91 92 28 59 66 84 97 97
f 10 18 45 65 83 91 8 39 43 64 77 78
g 5 16 16 30 59 62 84 84 90 97
h 48 51 56 59 71 12 45 47 82 95
i 10 15 37 48 59 91 24 37 52 68 69 84
j 0 57 57 66 74 36 63 63 73 83
k 14 19 43 46 58 60 30 37 50 53 90 99
l 25 29 33 49 59 98 6 26 59 60 68 93
m 9 65 65 74 88 12 64 64 66 77
n 14 14 40 40 98 44 44 73 73 93
o 32 46 50 63 70 80 20 32 54 89 93 94
p 4 48 48 51 92 23 40 41 74 88 -
select from t where 16<=(count;AskOrder) fby stock
You’d want an fby for that filtering