davidcrossey
Forum Replies Created
-
I’m not certain, but using a simple example, column order in .d file doesn’t appear to impact the operation of .Q.dpft (which is called by .Q.hdpf):
q)get hsym `$"./2021.11.25/t/.d" `sym`qty`price q)get hsym `$"./2021.11.26/t/.d" `sym`price`qty q)\l . q)t date sym price qty ---------------------------- 2021.11.25 akjl 90.30751 2 2021.11.25 amge 91.49882 33 2021.11.26 jobl 64.2737 39 2021.11.26 lfjp 38.69818 48
One can check through the definition of .Q.hdpf to see what it’s doing; granted it’s implemented in k), however you can see some clear parts:
q).Q.hdpf k){[h;d;p;f](@[`.;;0#]dpft[d;p;f]@)'t@>(#.:)'t:.q.tables`.;if[h:@[hopen;h;0];h"\l .";>h]}which as per link in my initial response aligns to this statement of 3 parts:
“Saves all tables by calling .Q.dpft, clears tables, and sends reload message to HDB.”
The question to ask yourself is where is it failing?
If the data is being saved to your HDB then it’s getting past the .Q.dpft call. The type error could be coming from the reload, where you RDB is trying to reload the HDB via handle ‘h’. Are you passing an HDB handle to your process as the first parameter, is that handle valid etc
-
Hi k-mambo,
A type error is a very general error, meaning it could be any number of issues.
The best place to start would be to confirm what arguments you are passing to .Q.hdpf and that each of the parameters are of the correct type as per .Q.hdpf
.Q.hdpf[historicalport;directory;partition;`p#field]
Additional resources:
- See ‘End of day‘ here: Building real-time tick subscribers | White Papers | kdb+ and q documentation – Kdb+ and q documentation (kx.com)
- See .Q.dpft (called by .Q.hdpf) Building real-time tick subscribers | White Papers | kdb+ and q documentation – Kdb+ and q documentation (kx.com)
Thanks for sharing your question on the KX Community!
Hope this helps,
David
-
Hi powerpeanuts,
I would keep it simple as you have it, and change select to update in your example as follows:
q)update change:0^price-prev[price] by sym from trade time sym price size change ----------------------------------- 09:30:00.000 a 10.75 100 0 09:31:00.000 a 11.75 100 1 09:32:00.000 a 13.2 100 1.45 09:30:00.000 b 100.75 100 0 09:31:00.000 b 106.95 100 6.2 09:32:00.000 b 123.95 100 17
Note, I’ve also used fill (^) here to set the first change value for each sym as 0; this is optional and you can leave it as null if you prefer.
Hope this helps and answers your question.
Kind regards,
David
-
-
Hi powerpeanuts,
You would need to add your new column first, before trying to use it in the where clause. For example:
q)t:([] name:`Dent`Beeblebrox`Prefect; iq:98 42 126) q)select name, iq2 from (update iq2:iq%100 from t) where iq2>1 name iq2 ------------ Prefect 1.26
Thanks for posting your question on the KX Community. Great username too!
Kind regards,
David
-
Hi codh,
Thanks for your query on the KX Community!
@kjame is spot on here .Q namespace tool usage is encouraged as it provides a lot of useful functionality. Users should only use the publicly documented API as per The .Q namespace.
Undocumented functions are subject to change without warning or notice as per the disclaimer in .q.k – usage is at the user’s own risk.
We can look into documenting changes to the publicly documented API as these occur.
Kind regards,
David
-
Hi Leah,
A kdb logfile typically has the form of (func;tabName;tabData), which can be as simple as f:{upsert[x;y]}. There isn’t anything special in terms of replaying the data into the kdb process, but more so to do with saving down the data.
Assuming you mean to save to another date, you could simply pull out a date from the data (after replay) and use it as a parameter for your save down logic.
If you mean at a higher level on partitioning the data differently across the entire database, data can be partitioned by any column of types date, month, year and int (1, 2), so depending on the requirements you could map values to a specific int based partition schema (3).
Hope this helps!
Kind regards,
David
References:
-
Hi planefan,
Radians tend to be the more natural unit when working with angles than degrees, and you’ll find they are typically the default unit in most languages.
KDB doesn’t have built in support for degrees, however you can work in degrees by converting to radians and rounding your results.
For example:
q)deg:0,0+12#30 q)deg 0 30 60 90 120 150 180 210 240 270 300 330 360 q)pi:3.141592653589793 q)round:{(10 xexp neg x)*`long$y*10 xexp x}[3;] /3 dp; change as necessary q)deg!round sin deg*pi%180 0 | 0 30 | 0.5 60 | 0.866 90 | 1 120| 0.866 150| 0.5 180| 0 210| -0.5 240| -0.866 270| -1 300| -0.866 330| -0.5 360| 0You can set the above as a function for easier re-use; and also swap sin, for cos or tan, although be careful with tan(90) and tan(270) which should be be set to null / undefined.
Kind regards,
David
-
Hi planefan,
Thanks for posting your query on KX Community!
For the angles you are trying to solve, are you working in degrees or radians?
KDB assumes a radian input, see here using sin as an example. You can see how to convert between degrees and radians on Geometry and trigonometry
Kind regards,
David
-
Hi MEdan,
It depends if your variable is a dictionary or a table as this flips the indexes:
- dictionary = d[key;index]
- table = t[index;key]
For example:
q)d:`name`iq!(`Dent`Beeblebrox`Prefect;98 42 126) q)d name| Dent Beeblebrox Prefect iq | 98 42 126 q)t:flip d q)t name iq -------------- Dent 98 Beeblebrox 42 Prefect 126 q)d[;2] name| `Prefect iq | 126 q)d[`name;] `Dent`Beeblebrox`Prefect q)d[`name;2] `Prefect q)t[2;] name| `Prefect iq | 126 q)t[;`name] `Dent`Beeblebrox`Prefect q)t[2;`name] `Prefect
Passing only the symbol to either dictionary or table will return the associated values key/column:
q)d `iq 98 42 126 q)t `iq 98 42 126
Further reading:
- Tables | Q For Mortals | A textbook forkdb+ and the q programming language – Q for Mortals (kx.com)
- 5. Dictionaries – Q for Mortals (kx.com)
Thanks for posting your query on the portal!
Kind regards,
David
-
Hi Simon,
Please find some insight from our team as follows;
embedpy defaults to numpy arrays because that makes a lot more sense for the typical usage pattern.
numpy arrays behave enough like lists that this is usually not a problem.
In a rare case that it is, it’s reasonable to do explicit conversion by calling list(). it’s exposed to q as .p.list, so e.g. to pass a float vector to print() as a python list instead of a numpy array,
q).p.print .p.list 1 2 3f [1.0, 2.0, 3.0]
That’s exactly the same as doing that in python, but saves defining a function
q).p.eval["lambda x:print(list(x))";1 2 3f] [1.0, 2.0, 3.0]
Thanks for adding your question on the portal!
Kind regards,
David
-
Hi ekallivrousis,
The largest decimal value would be the float type as per Data types
q)(2 xexp 1023)*1+1-2 xexp -52 1.7976931348623157e+308
See here for more Double-precision floating-point format
Also note that the q console can only display float values to 17 significant figures, e.g.
q)P 7i q)1.123456789 1.123457 q)P 0 /this is eqiuvalent to P 17 q)P 0i q)0.11111111111111112 0.11111111111111112 q)1.11111111111111112 1.1111111111111112
Further reading here:
Hope this helps, and thanks for sharing your query on the Community portal!
Kind regards,
David
-
Hi @ekallivrousis,
You could use the Double / Float type as per the C API Reference
Thanks for posting your query on the forums!
Kind regards,
David
-
Hi Simon,
Thanks for raising this query on the Community portal, and apologies in the delayed response.
Regarding the KX Platform installer, ARM 64-bit isn’t defined as it’s not a supported architecture at this time.
We do have a ARM 64-bit version of kdb for Linux and macOS for v4.1t that can be downloaded from downloads.kx.com – note this is a test version and is not intended, nor supported for production usage.
Hope this helps.
Kind regards,
David