david_navigator

odd, same link takes me to

Odd. I just get a page with a load of flags. Maybe you need a login to get a price ?

Looks great. Any idea of the cost as the web site doesn't even give a clue 😞

What does Text Tools | Comment Member(s) do ? I can't see it in the documentation. Thanks David

Ah thanks. I was looking for something to comment out the whole of a class in one go.

Not sure if something has changed in 10.4.2 but the following code works under 10.3 but not 10.4.2. The problem is the line move(temp[1],Buffer^, length(temp) * sizeof(WideChar)); which raises an exception if LoginMessage is empty. Stepping through, temp = #0 which I'm guessing is why move(temp[1].... fails, but why would it have worked previously (different compiler settings ) ? procedure TnxdsrPluginCmdHandler.nmLoginMessage(var aMsg: TnxDataMessage); var error: TnxResult; Buffer: PWideChar; BufLen: SmallInt; temp: WideString; reply: TnxnmServerlogonRpy; begin BufLen := high(Reply.Buffer) + 1; Buffer :=AllocMem(BufLen); try if assigned(PluginEngine) then temp := WideString(TnxdsrServerPlugin(PluginEngine).LoginMessage) else temp := ''; move(temp[1],Buffer^, length(temp) * sizeof(WideChar)); error := 0; BufLen := Length(Buffer); reply.buflen := buflen; copymemory(@reply.buffer, buffer, Buflen * sizeof(WideChar)); assert(buflen = length(buffer), 'Data Checking data Lengths differ'); except reply.BufLen := 0; error := DBIERR_NOSERVERSW; end; Cheers David

Actually it's code buried deep in a third party component and I only looked at it because it started to error with D10.4.2 As I said, it worked (well at least functioned) correctly for the past 10 years under various versions of Delphi (up to 10.3) >Why use Move and CopyMemory? They do the same job. Pick one and use it consistently. I thought one dealt with overlapped memory & one didn't ? Though no idea why the original code used both. I guess the 2 min fix to test that Test is not an empty string is going to turn in to a major rewrite.

Absolutely. I can see why the code fails and how to fix it, what I don't understand is why it's been working for 10 years. This routine is called 100's of times a day and Temp is virtually always an empty string, but the exception only started when I compiled the app with D 10.4.2, so I suppose my question is, why ? Maybe there's a compiler setting which has a different default in 10.4.2 ?

Is there a better way to get the list of Windows sounds as displayed in this Sounds dropdown without simply enumerating the contents of the %windir%\Media folder ?

Nothing. I just didn't want to do that IF there was an API that was the prefered method, especially as I couldn't find any documentation as to whether the folder name is localized in none English Windows.

I need to store some data indexed on two physical street addresses - a source and a destination. Sometimes this street address might be as simple as just a Country, other times it could be a full Address down to the building's number. I don't need to actually store the addresses. In reality the source address will probably only have about a dozen different values, whereas the destination address will run in to 1000's. I'm guessing that some kind of HASH is required of each address but this is an area that I don't really know much about and after any advice.

I think it's the sharing effect. I can't think of the number of times I've described an issue in a NG or a forum and realised the solution before I pressed submit.

Thanks for all the replies and suggestions. Just went for a walk with the dog and realised that I don't need the address info at all, I can index the other data fields (5 x integers) to find the record I need 🙂 Amazing how doing something different often brings alternative solutions to the forefront.

Sadly not. In the USA Zip codes are not as precise as Netherlands or UK, as they refer to the sorting & delivery district, not the street.

Yes. The system that sends the address data is pretty good at making sure that there isn't more than one version of an address for a physical location and for those that did slip though it would be acceptable to have "duplicate" data in this table.

In this particular scenario '1688 High St.' & '1688 High Street' would represent two distinctly different locations and thus two different records in the database. Even '1688 High St' (no period) would represent a third location.

I think I'm using the term HASH to refer to a checksum and was kind of hoping that different addresses would give different checksums. Kind of how What3Words works.

Maybe I'm using the wrong terminology - this isn't a area I'm too clued up on. I was hoping that irrespective of the length of the address I could calculate some value (of known max length) that uniquely represented the address and could be used in in the database table e.g in pseudo code UNIQUEVALUE1 := Function ConvertAddress('1688highstknowlesolihullb930ly'); UNIQUEVALUE2 := Function ConvertAddress('1600pennsylvaniaavenuenwwashingtondc20500unitedstates'); DB TABLE UNIQUEVALUE1 | UNIQUEVALUE2 | Float 1 | Float 2 | etc

Not sure I follow you. I have two addresses and against those I need to store about 10 floats. This will all be in a database table. The two addresses come from a third party app, they're sent as structured string data e.g Field1 - address_line_1 Field2 - address_line_2 Field3 - Town Fied4 - County Field5 - PostCode Field6 - Country Fields can be empty. There's no unique ID or anything useful, just a load of strings. The one thing that is consistent is that the same street address will always have the same field values. Knowing the original source of the data, I think it's acceptable to concatenate all the fields into one string and remove all whitespace/punctuation. Thus I could be sent Field1 - 1688 High St. Field3 - Knowle Fied4 - Solihull Field5 - B93 0LY and convert it to 1688highstknowlesolihullb930ly which I could then store as one of the keys e.g Source | Destination | Float1 | Float 2 | etc 1688highstknowlesolihullb930ly | 1600pennsylvaniaavenuenwwashingtondc20500unitedstates | Float 1 | Float 2 | etc But as there's no maximum length of any of the field data I'm sent, these keys could end up being unnecessarily long, so this doesn't feel like the right approach.

I just discovered a nasty bug in my code. During a previous debugging session I'd commented out the use of a parameter and replaced it with a hardcoded value i.e aPort replaced with 16000 procedure TAPI_Handler.ConnectDatabase(const aIPaddress: string = '192.168.42.212'; const aPort: Integer = 160000); ... ... ... FnxWinsockTransport.Port := 16000; // aPort; Surprisingly the compiler doesn't issue a hint or a warning that aPort isn't used, neither does Pascal Analyzer. Does anyone know of any third party code checking tool (maybe TMS ?) that would have picked this up ? Cheers David

Thanks.... and had I declared it as a Word, then the compiler would have picked up that I'd accidentally added an extra zero in to the default value (160000, rather than 16000) and I wouldn't have needed to debug and thus create my own bug !!!

I just compiled a colleagues code with Range checking turned on and got an immediate error. It's fairly obvious why and an easy fix (aLen had a value of 9 in this circumstance and the bytes array is simply the wrong size), but I was wondering what actually happens when the faulty code is executing with range checking turned OFF ? What actually happens when the line abyte := bytes[B]; is executed with a value of 8 for B ? Is there a potential for memory corruption or does the code internally fail gracefully ? function GetSetCardinality(const aSet; aLen: Word): Word; var B, i: Integer; abyte: Byte; bytes: array [0 .. 7] of Byte absolute aSet; begin result := 0; for B := 0 to aLen - 1 do begin abyte := bytes[B]; ... ... Many thanks Cheers David

8 different types.

Quickly Grepping the source, it looks like there are 8 independent sets that use the GetSetCardinality function.

Brilliant. Thank you very much. Saved me hours of googling & head scratching 🙂 Happy New Year.