Coding Horror

Slutmonkey57b · 02-10-2020 8:56pm

iLikeWaffles wrote: »

It is a quote from the Minister of Education talking to the press. So are you saying that she is wrong about the whole thing?

A politician would never lie.

ShatterProof · 02-10-2020 8:59pm

Slutmonkey57b wrote: »

A politician would never lie.

100% of the time

beauf · 02-10-2020 9:06pm

The politics of it don't interest me.

I'm just curious about it from a development point of view, and how they deal with situation.

Besides which, there are no bugs, just undocumented features.

iLikeWaffles · 02-10-2020 9:40pm

beauf wrote: »

I never said she was wrong ; )

It was a question not a statement!

GreeBo · 03-10-2020 1:26pm

Wombatman wrote: »

On the contrary this is a very easy case for testing as it is numbers based solution and isn't complicated by the vagaries of user behavior.

The calculations could have absolutely been done on paper to validate for a sample control set of individual students.

It's only numbers based if the requirements provide all the input scenarios and all the expected results.

The problem here is that the system itself is the thing providing expected results so only a code review could find the issue.

Wombatman · 03-10-2020 8:16pm

Must be riddled with bugs.......

The third error discovered in this latest review had to do with the incorrect treatment of Junior Certificate results belonging to students who were missing one of the three core subjects of Irish, English, and Maths. The company contracted to carry out this latest review Educational Training Systems corrected this newly discovered mistake during the review process.

ETS also found a fourth "issue" which "does not exactly match what is described". However the Department was reluctant to call this an error. It said it was an anomaly which had not impacted on students' results.

https://www.rte.ie/news/education/2020/1003/1169207-leaving-cert-calculated-grades/

beauf · 03-10-2020 9:09pm

Its the nature of software development to have bugs. Nothing remarkable about that.

iLikeWaffles · 03-10-2020 10:19pm

beauf wrote: »

Its the nature of software development to have bugs. Nothing remarkable about that.

It's the nature of software developers to find bugs.

Treppen · 03-10-2020 10:44pm

I heard that they are going to publish the code!

Does anyone know how the JC results exactly impacted on the LC results. All I could find is this.

The model was designed to take the Junior Cycle results of a group of students and use that group’s results to inform the calculation of their Leaving Certificate Results. It bears re-stating; the system did not take the results of a single student’s Junior Cycle exams and apply it to that specific student’s Leaving Certificate. Rather, the system looked at performance at an aggregate class level in the Junior Cycle examinations, and applied that in calculating students’ Leaving Certificate Results.

https://www.gov.ie/en/press-release/bc698-statement-from-the-minister-for-education-norma-foley-td-and-the-department-of-education-and-skills-regarding-leaving-certificate-2020-calculated-grades-process/

beauf · 03-10-2020 10:45pm

iLikeWaffles wrote: »

It's the nature of software developers to find bugs.

I though that was live testers (users)

The Continental Op · 03-10-2020 10:50pm

I've just found an interesting bug in APCs PowerChute Personal Edition software. I thought something was desperately going wrong until I did a reboot and the error went away.

I should have taken screen shots but according to PowerChute Personal Edition my input (mains) voltage was varying between 230 Volts and roughly 197000000 Volts.

Oh well what do I expect of free software.

beauf · 03-10-2020 10:55pm

That dept info could be interpreted in a number of ways, one could be it chooses a different calculation (unlikely), the other could it changes some value that the calculation uses.

Imagine you were in a bad class for JC but did well and got moved up to better classes for LC.

Slutmonkey57b · 03-10-2020 10:56pm

That doesn't make any sense to normalise results on an aggregate basis. The whole point of individual achievement is its particular to the individual

nthclare · 03-10-2020 10:58pm

My son's started studying coding, bless him his dad's a horticulturist.
Although my brother is a coding guy, and it's tough going he says

beauf · 03-10-2020 10:59pm

Slutmonkey57b wrote: »

That doesn't make any sense to normalise results on an aggregate basis. The whole point of individual achievement is its particular to the individual

Seems odd doesn't it.

Earthhorse · 03-10-2020 11:08pm

Especially considering the make up of a class can change over time. In fact the end of Junior Cycle I would say is when most transfers take place between schools. But if there’s a correlation maybe it’s right for them to include it in the algorithm? Don’t really know enough about the area to say.

Treppen · 04-10-2020 9:17am

Slutmonkey57b wrote: »

That doesn't make any sense to normalise results on an aggregate basis. The whole point of individual achievement is its particular to the individual

Ya it's all a bit odd. Considering you could have maybe 140 students in some schools, all doing various combinations of 10+ subjects in JC then moving to various 7 subjects in the LC (with dropouts, moves to other schools, exemptions in Irish.. Often those subjects had little connection to JC. E.g. JC science to LC physics or applied maths.

And then on top of that say that the error only impacted on one LC subject!

Anyway, code is going to be released apparently.

Wombatman · 05-10-2020 10:04am

Awarding lower grades than should have been given only half the problem.

Almost 8,000 Leaving Cert grades issued this year were higher than they should have been due to coding errors in the calculated grades process, according to new figures.

In addition to Saturday’s announcement that 6,100 students received lower grades than they should have, the Department of Education confirmed on Sunday evening that about 7,943 grades were higher than they were supposed to be.

While lower grades will be corrected upwards, higher grades awarded in error will not be corrected.

Capt'n Midnight · 05-10-2020 10:37pm

https://www.theguardian.com/politics/2020/oct/05/how-excel-may-have-caused-loss-of-16000-covid-tests-in-england

In this case, the Guardian understands, one lab had sent its daily test report to PHE in the form of a CSV file – the simplest possible database format, just a list of values separated by commas. That report was then loaded into Microsoft Excel, and the new tests at the bottom were added to the main database.

Idleater · 06-10-2020 7:52am

Capt'n Midnight wrote: »

https://www.theguardian.com/politics/2020/oct/05/how-excel-may-have-caused-loss-of-16000-covid-tests-in-england

I saw a tweet about that yesterday, the "fix" was to start using multiple spreadsheets.

06-10-2020 8:01am

Wombatman wrote: »

Awarding lower grades than should have been given only half the problem.

While lower grades will be corrected upwards, higher grades awarded in error will not be corrected.

Bit harder to do given that those that received them have most likely already started their college courses?

John_C · 20-10-2020 4:41pm

I found an entertaining bug today. It's a function to take in a duration in seconds and return a text string for display.
For example getDisplayDuration(123) should return 02m 03s and getDisplayDuration(3723) should return 01h 02m 03s.
Instead of working out the mathematics, the function converts the duration into a Date object, i.e. what time was it this number of seconds past midnight on 1st January 1970.
It then reads the hours, minutes and seconds from that object.
The end result is that (assuming you live in Ireland) the function produces the correct output during winter time but is off by an hour during daylight savings time.

function getDisplayDuration(duration) {
  if (duration >= 60 * 60) {
    var d = new Date(duration * 1000);
    return (
      addZero(d.getHours()) +
      "h " +
      addZero(d.getMinutes()) +
      "m " +
      addZero(d.getSeconds()) +
      "s"
    );
  } else {
    var d = new Date(duration * 1000);
    return addZero(d.getMinutes()) + "m " + addZero(d.getSeconds()) + "s";
  }
  function addZero(i) {
    if (i < 10) {
      i = "0" + i;
    }
    return i;
  }
}

22-10-2020 8:38am

John_C wrote: »

I found an entertaining bug today. It's a function to take in a duration in seconds and return a text string for display.
For example getDisplayDuration(123) should return 02m 03s and getDisplayDuration(3723) should return 01h 02m 03s.
Instead of working out the mathematics, the function converts the duration into a Date object, i.e. what time was it this number of seconds past midnight on 1st January 1970.
It then reads the hours, minutes and seconds from that object.
The end result is that (assuming you live in Ireland) the function produces the correct output during winter time but is off by an hour during daylight savings time.

Reminds me of this...

Idleater · 22-10-2020 8:57am

John_C wrote: »

The end result is that (assuming you live in Ireland) the function produces the correct output during winter time but is off by an hour during daylight savings time.

I've seen a publishing schedule done by time of day, but specified by an arbitrary date. Basically "11:30 daily", but specified as "11:30 22 October 1950 ET".

I don't recall the specifics, but the rigmarole to calculate "today" based on where the user is (say GMT, or more fun, Asia) :mad:

Wombatman · 22-10-2020 9:21am

From the kind of developers that brought you 'We fooked up importing the German test results" and "We made a balls of the leaving cert standardisation process"..........

https://arstechnica.com/tech-policy/2020/10/excel-glitch-may-have-caused-uk-to-underreport-covid-19-cases-by-15841/

dazberry · 22-10-2020 10:40am

Idleater wrote: »

I've seen a publishing schedule done by time of day, but specified by an arbitrary date. Basically "11:30 daily", but specified as "11:30 22 October 1950 ET".

This reminds me of a place I worked where the empty/null/not supplied date for a customer dob was some arbitrary date in 1966. Caused a problem a few times

Capt'n Midnight · 22-10-2020 5:47pm

dazberry wrote: »

This reminds me of a place I worked where the empty/null/not supplied date for a customer dob was some arbitrary date in 1966. Caused a problem a few times

In the US a few lifers were released on the 9th of September back in 1999

Slutmonkey57b · 22-10-2020 8:27pm

Wombatman wrote: »

From the kind of developers that brought you 'We fooked up importing the German test results" and "We made a balls of the leaving cert standardisation process"..........

https://arstechnica.com/tech-policy/2020/10/excel-glitch-may-have-caused-uk-to-underreport-covid-19-cases-by-15841/

Over a million rows in excel. I bet they went off and made dinner waiting for the thing to open.

Seth Brundle · 23-10-2020 4:55pm

Slutmonkey57b wrote: »

Over a million rows in excel. I bet they went off and made dinner waiting for the thing to open.

You're assuming that they're using the newer Excel formats availabel in Excel 2007 onwards.
It wouldn't surprise me to see them using the *.xls format which is limited to 65536 rows.

Earthhorse · 23-10-2020 4:58pm

That's what the issue was according to the article.

sqlmonkey · 23-10-2020 5:18pm

ah yes - the old Excel Max Rows gotcha!

Seth Brundle · 23-10-2020 5:27pm

sqlmonkey wrote: »

ah yes - the old Excel Max Rows gotcha!

If you're familiar with Excel macros then it's a basic check. Surprised it happened but then again, if they were using Excel instead of an actual database (even MS Access) then they deserve all the criticism they get!

Buford T Justice · 24-10-2020 1:45am

Access is (was) limited at 50MB

Seth Brundle · 24-10-2020 11:30am

Buford T Justice wrote: »

Access is (was) limited at 50MB

I'm fairly sure that the max size for an Access 97 database is 1GB. Access 2000 onwards is 2GB.
I can recall seeing desktop applications using Access macros which were several hundred megs.
I'm not saying it's perfect but it's not limited to 50Mb

beauf · 24-10-2020 12:49pm

I used to have run a Job that maxed out an access database. I think it was 2GB. Every month had to compact it, which would free up some space which would allow me to run the job. Took two weeks to run, and if it failed, I barely had enough time to rerun it before month end. They wouldn't let me rewrite it or port to SQL. Through lateral thinking I eventually got it to run in 24hrs.

Usually when they get that big you throw the data into SQL and keep the front end in Access via linked tables. Because there's usually a mountain if VBA connected to it, you don't want to touch.

Capt'n Midnight · 24-10-2020 3:15pm

beauf wrote: »

Because there's usually a mountain if VBA connected to it, you don't want to touch.

*shudders*

Printout worked on a 386 but not on the 486 that replaced it

There was a delay loop in the code. I increased the counter and left notes.

dazberry · 24-10-2020 4:29pm

Capt'n Midnight wrote: »

*shudders*

Printout worked on a 386 but not on the 486 that replaced it

There was a delay loop in the code. I increased the counter and left notes.

Reminds me of the Turbo/Borland Pascal CRT unit timing bug.

jhegarty · 24-10-2020 4:33pm

Seth Brundle wrote: »

I'm fairly sure that the max size for an Access 97 database is 1GB. Access 2000 onwards is 2GB.
I can recall seeing desktop applications using Access macros which were several hundred megs.
I'm not saying it's perfect but it's not limited to 50Mb

I seem to remember the free local only version of SQL Server was limited to 50mb. That might be what the poster is thinking of.

27-10-2020 2:25pm

beauf wrote: »

Through lateral thinking I eventually got it to run in 24hrs.

You can't just leave us hanging. :pac:

What did you do?

beauf · 28-10-2020 12:03am

Deleted User wrote: »

You can't just leave us hanging. :pac:

What did you do?

The simplistic answer is I split the data into to smaller chunks, ran it on multiple PCs concurrently then re-joined it together at the end But I got it down from about 2wks to about 3~4 days doing this. But it was still a PITA. Eventually I lost the use of the extra machines. So I replaced them all with a single overlocked early i3 @4ghz which chewed through it in about 24 hours. Access being single threaded ran much faster on high clock speeds than on our dual CPU Xeons of the day.

In another place I saw a similar solution. Big bloated Oracle system, ran like a dog. One of the customers got fed up waiting for developers to optimize it. Bought some crazy sever with like 16+ cpu's (a lot at that time), and it ran perfectly on that.

Wombatman · 02-11-2020 10:39pm

AlmightyCushion · 03-11-2020 2:22am

Infinite loop. Rookie mistake.

meep · 04-11-2020 1:58pm

AlmightyCushion wrote: »

Infinite loop. Rookie mistake.

GetTested() will have HCF if it fails.

fcrossen · 04-11-2020 6:23pm

var is never reassigned - use const instead.

iLikeWaffles · 04-11-2020 7:19pm

fcrossen wrote: »

var is never reassigned - use const instead.

What so have a load of dead people getting tested, wearing masks, 6 feet apart and washing hands?!

Seth Brundle · 04-11-2020 8:10pm

iLikeWaffles wrote: »

fcrossen wrote: »

var is never reassigned - use const instead.

What so have a load of dead people getting tested, wearing masks, 6 feet apart and washing hands?!

Surely those who don't wear masks, don't keep 6 feet apart or don't wash hands are the const?

iLikeWaffles · 04-11-2020 9:00pm

Would not be able to change a const which is true to false!

var dead = !still_alive; &#55358;&#56614;*♀️

spaceCreated · 15-11-2020 9:49pm

Checks this thread... still don't see any of my code... relax

iLikeWaffles · 15-11-2020 10:51pm

spaceCreated wrote: »

Checks this thread... still don't see any of my code... relax

You need to post it first :pac:

Oink · 15-11-2020 10:59pm

AlmightyCushion wrote: »

Infinite loop. Rookie mistake.

I got my very first infinite loop in JavaScript last week.
I let it run for a bit to enjoy the moment.

Coding Horror

Comments