Next month: October 1996
The top-level of the UK and InterLotto trees were cluttered with a load of winning numbers pages, so I decided to move them into a sub-directory called "All", to match the link that's clicked on in the winning numbers index page. I did a complete internal link check for the UK and InterLotto pages after this of course and duly adjusted the links on various pages to point to the new location.
I've started the coding for a new section, which will entail a split of the sales figures (currently seven links at the bottom of the Numerical Analysis) into their own pages. Needless to say, there will be further page splits within the new section in the same style as the winning numbers section (i.e. 1994-1996 and All, pre-formatted and HTML tables). I make that a total of 51 pages (including the index page) in the new section ! I'm intending to take Wednesday and Thursday off to complete this coding and hopefully you'll see the results on Friday.
Unfortunately, none of them agreed with each other about the jackpot amount :-(
If you cast your mind back to 29th August, you'll remember that I bought my dad an 8-week ticket containing two lines per week for his birthday (yes, a pretty sad present if you ask me). Well, the second line came up trumps with a £10 win, which at least means the present wasn't a total loss.
Whilst "cleaning up" common definitions, I also decided to process the manual comments section of each individual lottery page through bhs. This has several benefits such as removing commented out HTML (I can use #if 0 instead) and the use of many pre-defined macros and definitions, which finally gave me the ability to easily put and icons into the individual lottery pages (which I'd not been able to do before).
I've left the manual section of each InterLotto individual lottery page unmacroised though, because I don't actually write that HTML for them of course (it's updated by InterLotto via quite a sophisticated system I wrote) and I don't want to have to teach the folks at InterLotto how to use bhs macros...
Discovered that my local Tesco store (a big one) had been selling lottery tickets since last Monday (after I quizzed the counter staff about it). It turns out the store, the biggest for several miles, had applied for a lottery terminal before the lottery began, but had been put on a waiting list ever since ! For the incredible 22 months that Tesco had been denied a terminal, a tiny newsagent in the town centre was the main outlet, which is sheer madness on Camelot's part really.
Tidied up the Notes section at the bottom of tables in the winning numbers section. I'd mentioned ball sets on pages that didn't list them (i.e. all but the draw order pages) and also referred to sets A and B for draw order pages that didn't list them (i.e. any that only covered draws after Lottery #8).
Internet Explorer 3.X has an annoying bug that wasn't present in 2.X: it now line-by-line centres pre-formatted text following a <CENTER> or <P ALIGN=center> tag (I use the latter) ! This means that any "short" lines are centred differently from longer ones and this hit the pre-formatted number grid in the Number Palace. To work around this bug, I padded the last line of the grid with extra spaces to make all lines the same byte length.
The Camelot fax didn't come through, so I phoned them up and got all the mid-week figures except the Lucky Dip, which I'll presumably be able to get tomorrow.
Without announcing it publicly, we'd switched off DNS lookup on the secure (https) port of the Apache server we run on the Connect machine about a week ago. It appears that when you approach about an average of one hit a second (which is where we're at now), hostname lookup via DNS noticeably impacts server performance. Turning it off means that any CGIs only get passed the numeric IP address instead of the fully qualified host name, which meant I had to adjust the virtual lottery code to compare against numeric domains as well as named ones (not difficult).
Anyway, to cut a long story short, we switched off DNS lookup for the insecure (http) port today as well, which should hopefully speed things up for people accessing the WWW server because we won't be looking up their hostnames for logging purposes. Of course, I also had to get the newest version of analog, the WWW log file analyser we use, to use its own DNS cache file so that logs could be processed with a translation of numeric addresses in the logs into full addresses.
Spotted an interesting (if rather one-sided) interview in a Student Newsletter (dated December 1995) with a Camelot representative that's available on the WWW. They actually admit that they get the lottery results (number of winners etc.) "immediately" after the 7th ball appears, so why on earth can't they present that on the TV show as well ? Mind you, confusingly, the Camelot spokesperson says "by nine o'clock..." and "the process is completed in 45 minutes", neither of which is "immediately" or "almost instantly" in my books !
The interview also failed to mention that it is possible for retailers to cancel a recently purchased ticket and pocket the money (i.e. there was no chase-up of the answers to "Are there any problems using the terminals from the players' or retailers' points of view ?" or "How confident is Camelot about the security systems ?").
It was complicated by the fact that I had to create 49 new bitmaps with the numbers and the words "Bonus Ball" in them. All these new bitmaps had to be the same width as the balls without the extra bonus text, so I had to move the bonus ball 8 pixels closer to the other 6 balls. The gap between the 6th and 7th ball is still double the others because I pad the main number balls on the right and the bonus ball on the left.
More "fun" was to come because I also had to generate 5 normal and 5 bonus balls, each with a "?" in them and coloured differently. This was to allow me to generate the next lottery ball graphic with question marks instead of numbers. So this made a total of 59 new UK lottery ball bitmaps (in native ImageMagick miff format) and 5 new InterLotto ball bitmaps.
Believe you me, I nearly gave up coding this half-way through because my brain was starting to overheat, but I made it to the end and this will prevent last weekend's disaster with the faulty screen dump of the balls. Note that the graphical frequency barcharts are still screen dumped because there's no easy way to avoid doing so.
Slightly speeded up the winning number page re-generation by only generating this year's and the "all years" pages by default and only generating the rest (i.e. 1994's and 1995's figures) when I upload pages from my SyQuest (in case coding changes have meant a layout change to even these old pages). This should mean that Saturday evening cron jobs should complete about 10 seconds quicker.
Changed the virtual lottery code, FAQ page and Rules page to reflect the new way of paying prizes (into an InterLotto account and in Swiss Francs). I'll leave the old code in (conditionally compiled out) just in case I have to revert back in the future. I've included the pounds sterling figure in brackets (using a weekly exchange rate I pick up from teletext) whenever the SFr is mentioned on its own, otherwise people will no doubt ask me what the sterling figure is ! The only really tricky bit was the need to convert SFr to pounds in the CGI code for the virtual lottery Have You Won ? page to keep it consistent throughout and allow a grand total to be displayed.
The Good Causes grand total exceeded MAXINT and caused my floating point display code to go wrong (it casted to an int and subtracted to get the fractional part). I duly adjusted the code so that it continually subtracted MAXINT until the whole part was under MAXINT.
Previous month: August 1996