"If you can't beat them, join them" is what they say, and that's exactly what I've done in respect of Liberty BASIC! In an earlier blog post I compared BBC BASIC for Windows and Liberty BASIC, and explained my reasoning for believing BBC BASIC to be the better of the two. But that's not to say I can't see the attraction of LB to some people, particularly in having support for Windows GUI features as a standard part of the language rather than being provided in a library.
I've argued before, and still very much believe, that building in a rich set of GUI features as standard can result in a bloated language. This is especially true in the case of an interpreted language when every 'compiled' executable needs to include support for all those features, whether the user's program needs them or not. I've always suspected that this explains, at least in part, the huge size of Liberty BASIC executables (not far short of 3 Megabytes, however small the program).
But whilst the principle of providing GUI support by means of libraries (which is what BBC BASIC does) is sound, in practice there hasn't actually been support for the full range of facilities provided by Liberty BASIC, at least not in a straightforward way. Certainly there are BB4W libraries for the main Windows GUI elements, but LB goes somewhat further, for example you can change the colours of standard Windows controls (whether that's a good idea is arguable) and you can arrange that buttons are automatically relocated when a window is resized.
So I've recently remedied this by writing a BBC BASIC for Windows library LBLIB which makes available the full range of LB features to a BBC BASIC program. Indeed, it does so as far as possible in a way that is fully compatible with Liberty BASIC, and this has made possible a brand-new product: Liberty BASIC Booster!
But I'm getting ahead of myself in the story. Some while ago I wrote an automated translator from QBASIC (or Quick Basic) to BBC BASIC - QB2BBC. Unlike many translators this attempts to perform a 'perfect' translation that requires no user tweaking to achieve identical functionality to the original. The down-side of producing such an accurate translation is that the code it generates is often quite unlike what a human would write, and it doesn't give a good indication of what well-written BBC BASIC code looks like.
At the heart of QB2BBC is a parsing engine which breaks down the QBASIC program to its fundamental elements and then constructs the BBC BASIC program from them. It occurred to me some time ago that this parsing engine could be adapted quite easily to translate other languages into BBC BASIC, one obvious candidate being Liberty BASIC. So earlier this year I started to look quite seriously at the possibility of writing a Liberty BASIC to BBC BASIC translator (tentatively to be called LB2BBC).
But of course there was a big problem. With QBASIC originating from the 1980s, in the era of CP/M and MS-DOS, pretty much every feature it provides has a direct or near equivalent in BBC BASIC. Whilst not always straightforward, the task of QB2BBC was primarily one of translation rather than emulation (the only major exception being QBASIC's music capabilities, which are emulated at run-time by the QBLIB library). With Liberty BASIC however, translating the language syntax to BBC BASIC was only a small part of the challenge; it left the GUI facilties (principally windows, controls, graphics and sprites) completely unsupported.
This is where we complete the circle. By combining the translation capabilities of LB2BBC with the emulation capabilities of LBLIB it becomes possible to run most Liberty BASIC programs, without actually needing Liberty BASIC itself! In consultation with some leading Liberty BASIC advocates it became clear that there might be a market for this, not as as translator to BBC BASIC but as a 'Liberty BASIC Booster'. To the user it would appear as a way to speed up LB programs (up to ten times in some cases) and to create compact single-file executables, without it being obviously apparant that BBC BASIC was at work 'under the hood'.
The compatibility of LBB with LB isn't perfect. As I noted at the end of my article comparing the two BASICs there are two areas in which LB provides a capabilility that BBC BASIC doesn't: support for huge strings and arbitrary-precision integer arithmetic. Therefore programs relying on either of these features cannot run under LBB.
Some may find it superficially strange that I should be supporting Liberty BASIC in this way. One reason is that whilst LBB doesn't directly benefit BBC BASIC, neither does it significantly damage it: I wouldn't expect that any users of LBB might have been persuaded to convert to BB4W instead. Secondly, LBB demonstrates the amazing power and flexibility of BBC BASIC for Windows; with how many other high-level programming languages could it have been achieved? But mostly it's been a fascinating technical challenge, pushing my programming skills to the limit.
I've argued before, and still very much believe, that building in a rich set of GUI features as standard can result in a bloated language. This is especially true in the case of an interpreted language when every 'compiled' executable needs to include support for all those features, whether the user's program needs them or not. I've always suspected that this explains, at least in part, the huge size of Liberty BASIC executables (not far short of 3 Megabytes, however small the program).
But whilst the principle of providing GUI support by means of libraries (which is what BBC BASIC does) is sound, in practice there hasn't actually been support for the full range of facilities provided by Liberty BASIC, at least not in a straightforward way. Certainly there are BB4W libraries for the main Windows GUI elements, but LB goes somewhat further, for example you can change the colours of standard Windows controls (whether that's a good idea is arguable) and you can arrange that buttons are automatically relocated when a window is resized.
So I've recently remedied this by writing a BBC BASIC for Windows library LBLIB which makes available the full range of LB features to a BBC BASIC program. Indeed, it does so as far as possible in a way that is fully compatible with Liberty BASIC, and this has made possible a brand-new product: Liberty BASIC Booster!
But I'm getting ahead of myself in the story. Some while ago I wrote an automated translator from QBASIC (or Quick Basic) to BBC BASIC - QB2BBC. Unlike many translators this attempts to perform a 'perfect' translation that requires no user tweaking to achieve identical functionality to the original. The down-side of producing such an accurate translation is that the code it generates is often quite unlike what a human would write, and it doesn't give a good indication of what well-written BBC BASIC code looks like.
At the heart of QB2BBC is a parsing engine which breaks down the QBASIC program to its fundamental elements and then constructs the BBC BASIC program from them. It occurred to me some time ago that this parsing engine could be adapted quite easily to translate other languages into BBC BASIC, one obvious candidate being Liberty BASIC. So earlier this year I started to look quite seriously at the possibility of writing a Liberty BASIC to BBC BASIC translator (tentatively to be called LB2BBC).
But of course there was a big problem. With QBASIC originating from the 1980s, in the era of CP/M and MS-DOS, pretty much every feature it provides has a direct or near equivalent in BBC BASIC. Whilst not always straightforward, the task of QB2BBC was primarily one of translation rather than emulation (the only major exception being QBASIC's music capabilities, which are emulated at run-time by the QBLIB library). With Liberty BASIC however, translating the language syntax to BBC BASIC was only a small part of the challenge; it left the GUI facilties (principally windows, controls, graphics and sprites) completely unsupported.
This is where we complete the circle. By combining the translation capabilities of LB2BBC with the emulation capabilities of LBLIB it becomes possible to run most Liberty BASIC programs, without actually needing Liberty BASIC itself! In consultation with some leading Liberty BASIC advocates it became clear that there might be a market for this, not as as translator to BBC BASIC but as a 'Liberty BASIC Booster'. To the user it would appear as a way to speed up LB programs (up to ten times in some cases) and to create compact single-file executables, without it being obviously apparant that BBC BASIC was at work 'under the hood'.
The compatibility of LBB with LB isn't perfect. As I noted at the end of my article comparing the two BASICs there are two areas in which LB provides a capabilility that BBC BASIC doesn't: support for huge strings and arbitrary-precision integer arithmetic. Therefore programs relying on either of these features cannot run under LBB.
Some may find it superficially strange that I should be supporting Liberty BASIC in this way. One reason is that whilst LBB doesn't directly benefit BBC BASIC, neither does it significantly damage it: I wouldn't expect that any users of LBB might have been persuaded to convert to BB4W instead. Secondly, LBB demonstrates the amazing power and flexibility of BBC BASIC for Windows; with how many other high-level programming languages could it have been achieved? But mostly it's been a fascinating technical challenge, pushing my programming skills to the limit.
posted by Richard Russell | 2:15 PM
|
0 comments
