digitalmars.D - Pitching an investment bank on using D for their bond analytics
- D Denizen since a year (33/33) Apr 14 2015 Hi.
- Rikki Cattermole (4/36) Apr 14 2015 Just a thought, try getting them to use D for prototyping. Worse case
- Iain Buclaw via Digitalmars-d (4/51) Apr 14 2015 The best case scenario has happened at least once before in a
- D Denizen since a year (3/5) Apr 14 2015 Hi Iain.
- Iain Buclaw via Digitalmars-d (6/11) Apr 14 2015 My usual alias (Initial, Surname) at gdcproject.org - email is also
- bachmeier (3/71) Apr 14 2015 Or maybe they can write some smaller pieces in D and call them
- Rikki Cattermole (2/51) Apr 14 2015 Yeah, that would be middle ground.
- D Denizen since a year (4/7) Apr 14 2015 Good point. Pick a small part of the project (especially
- Rikki Cattermole (7/13) Apr 14 2015 That too, but I was thinking more along the lines of writing some
- D Denizen since a year (3/12) Apr 14 2015 Yes - the business logic is where I come in, so it makes sense
- Rikki Cattermole (2/12) Apr 14 2015 I'm available for hire if you need a developer/consultant btw.
- John Colvin (17/53) Apr 14 2015 A couple of big pluses:
- D Denizen since a year (8/23) Apr 14 2015 My guess is these guys aren't going to be writing much assembler,
- Daniel Murphy (7/9) Apr 14 2015 C++ integration can be made to work, as in data can be passed back and f...
- D Denizen since a year (8/18) Apr 14 2015 Is there anything anyone has on the gotchas for non-templated
- Benjamin Thaut (7/10) Apr 14 2015 You would have to asess if the analysis you should create is time
- D Denizen since a year (7/22) Apr 14 2015 My guess is it's not a major factor, but maybe can design from
- Dicebot (18/23) Apr 14 2015 From the business / management PoV I think it is important to
- Justin Whear (24/24) Apr 14 2015 On Tue, 14 Apr 2015 12:08:53 +0000, D Denizen since a year wrote:
- Andy Smith (69/78) Apr 15 2015 I've spent a few years in the trenches so here are a few
- Kapps (6/12) Apr 15 2015 One of the biggest issues I can think of would be code breakage.
- lobo (7/21) Apr 15 2015 In the last 2 years each breaking change I've encountered has
- ketmar (4/16) Apr 15 2015 not that anything is forcing someone to update the working compiler. i=2...
- Jacob Carlborg (4/8) Apr 15 2015 I don't agree. Every single release since at least DMD 2.050 has broken ...
Hi. I have been here a year or so, and trust you will forgive my posting pseudonymously on this occasion. If you guess who it is, please be kind enough not to say for now. A friend has been invited to be a consultant for an investment bank that would like to build a set of analytics for fixed income products. The team is currently quite small - about 5 C++ developers - and the idea is to start with a proof of concept and then build on it as there is further buy-in from the business. Having been using D for a year or so, I am pretty comfortable that it can do the job, and likely much better than the C++ route for all the normal reasons. I haven't experience of using D in a proper enterprise environment, but I think this group might be open to trying D and that I might be at least part-time involved. I also have little experience in getting across the merits of this technology to people very used to C++, and so I have not yet built up a standard set of answers to the normal objections to 'buying' that will crop up in any situation of this sort. So I am interested in: - what are the things to emphasize in building the case for trying D? the most effective factors that persuade people are not identical with the technically strongest reasons, because often one needs to see it before one gets it. - what are the likely pitfalls in the early days? - what are potential factors that might make D a bad choice in this scenario? I would like to use D certainly - but it is of course much more important that the client gets the best result, however it is done. - am I right in thinking C++ integration more or less works, except instantiating C++ templates from D? what are the gotchas? (I appreciate there is not so much to go on, and much depends on specific factors). But any quick thoughts and experiences would be very welcome.
Apr 14 2015
On 15/04/2015 12:08 a.m., D Denizen since a year wrote:Hi. I have been here a year or so, and trust you will forgive my posting pseudonymously on this occasion. If you guess who it is, please be kind enough not to say for now. A friend has been invited to be a consultant for an investment bank that would like to build a set of analytics for fixed income products. The team is currently quite small - about 5 C++ developers - and the idea is to start with a proof of concept and then build on it as there is further buy-in from the business. Having been using D for a year or so, I am pretty comfortable that it can do the job, and likely much better than the C++ route for all the normal reasons. I haven't experience of using D in a proper enterprise environment, but I think this group might be open to trying D and that I might be at least part-time involved. I also have little experience in getting across the merits of this technology to people very used to C++, and so I have not yet built up a standard set of answers to the normal objections to 'buying' that will crop up in any situation of this sort. So I am interested in: - what are the things to emphasize in building the case for trying D? the most effective factors that persuade people are not identical with the technically strongest reasons, because often one needs to see it before one gets it. - what are the likely pitfalls in the early days? - what are potential factors that might make D a bad choice in this scenario? I would like to use D certainly - but it is of course much more important that the client gets the best result, however it is done. - am I right in thinking C++ integration more or less works, except instantiating C++ templates from D? what are the gotchas? (I appreciate there is not so much to go on, and much depends on specific factors). But any quick thoughts and experiences would be very welcome.Just a thought, try getting them to use D for prototyping. Worse case they will play around with D before settling on e.g. c++. Best case scenario they'll move it into production.
Apr 14 2015
On 14 April 2015 at 14:16, Rikki Cattermole via Digitalmars-d <digitalmars-d puremagic.com> wrote:On 15/04/2015 12:08 a.m., D Denizen since a year wrote:The best case scenario has happened at least once before in a similarly-based company.Hi. I have been here a year or so, and trust you will forgive my posting pseudonymously on this occasion. If you guess who it is, please be kind enough not to say for now. A friend has been invited to be a consultant for an investment bank that would like to build a set of analytics for fixed income products. The team is currently quite small - about 5 C++ developers - and the idea is to start with a proof of concept and then build on it as there is further buy-in from the business. Having been using D for a year or so, I am pretty comfortable that it can do the job, and likely much better than the C++ route for all the normal reasons. I haven't experience of using D in a proper enterprise environment, but I think this group might be open to trying D and that I might be at least part-time involved. I also have little experience in getting across the merits of this technology to people very used to C++, and so I have not yet built up a standard set of answers to the normal objections to 'buying' that will crop up in any situation of this sort. So I am interested in: - what are the things to emphasize in building the case for trying D? the most effective factors that persuade people are not identical with the technically strongest reasons, because often one needs to see it before one gets it. - what are the likely pitfalls in the early days? - what are potential factors that might make D a bad choice in this scenario? I would like to use D certainly - but it is of course much more important that the client gets the best result, however it is done. - am I right in thinking C++ integration more or less works, except instantiating C++ templates from D? what are the gotchas? (I appreciate there is not so much to go on, and much depends on specific factors). But any quick thoughts and experiences would be very welcome.Just a thought, try getting them to use D for prototyping. Worse case they will play around with D before settling on e.g. c++. Best case scenario they'll move it into production.
Apr 14 2015
The best case scenario has happened at least once before in a similarly-based company.Hi Iain. What's best way to reach you by email for quick follow-up on this (if you would not mind - no hurry)?
Apr 14 2015
On 14 April 2015 at 16:21, D Denizen since a year via Digitalmars-d <digitalmars-d puremagic.com> wrote:My usual alias (Initial, Surname) at gdcproject.org - email is also listed on the site (it may take some looking). Regards Iain.The best case scenario has happened at least once before in a similarly-based company.Hi Iain. What's best way to reach you by email for quick follow-up on this (if you would not mind - no hurry)?
Apr 14 2015
On Tuesday, 14 April 2015 at 12:16:40 UTC, Rikki Cattermole wrote:On 15/04/2015 12:08 a.m., D Denizen since a year wrote:Or maybe they can write some smaller pieces in D and call them from C++.Hi. I have been here a year or so, and trust you will forgive my posting pseudonymously on this occasion. If you guess who it is, please be kind enough not to say for now. A friend has been invited to be a consultant for an investment bank that would like to build a set of analytics for fixed income products. The team is currently quite small - about 5 C++ developers - and the idea is to start with a proof of concept and then build on it as there is further buy-in from the business. Having been using D for a year or so, I am pretty comfortable that it can do the job, and likely much better than the C++ route for all the normal reasons. I haven't experience of using D in a proper enterprise environment, but I think this group might be open to trying D and that I might be at least part-time involved. I also have little experience in getting across the merits of this technology to people very used to C++, and so I have not yet built up a standard set of answers to the normal objections to 'buying' that will crop up in any situation of this sort. So I am interested in: - what are the things to emphasize in building the case for trying D? the most effective factors that persuade people are not identical with the technically strongest reasons, because often one needs to see it before one gets it. - what are the likely pitfalls in the early days? - what are potential factors that might make D a bad choice in this scenario? I would like to use D certainly - but it is of course much more important that the client gets the best result, however it is done. - am I right in thinking C++ integration more or less works, except instantiating C++ templates from D? what are the gotchas? (I appreciate there is not so much to go on, and much depends on specific factors). But any quick thoughts and experiences would be very welcome.Just a thought, try getting them to use D for prototyping. Worse case they will play around with D before settling on e.g. c++. Best case scenario they'll move it into production.
Apr 14 2015
On 15/04/2015 1:44 a.m., bachmeier wrote:On Tuesday, 14 April 2015 at 12:16:40 UTC, Rikki Cattermole wrote:Yeah, that would be middle ground.On 15/04/2015 12:08 a.m., D Denizen since a year wrote:Or maybe they can write some smaller pieces in D and call them from C++.Hi. I have been here a year or so, and trust you will forgive my posting pseudonymously on this occasion. If you guess who it is, please be kind enough not to say for now. A friend has been invited to be a consultant for an investment bank that would like to build a set of analytics for fixed income products. The team is currently quite small - about 5 C++ developers - and the idea is to start with a proof of concept and then build on it as there is further buy-in from the business. Having been using D for a year or so, I am pretty comfortable that it can do the job, and likely much better than the C++ route for all the normal reasons. I haven't experience of using D in a proper enterprise environment, but I think this group might be open to trying D and that I might be at least part-time involved. I also have little experience in getting across the merits of this technology to people very used to C++, and so I have not yet built up a standard set of answers to the normal objections to 'buying' that will crop up in any situation of this sort. So I am interested in: - what are the things to emphasize in building the case for trying D? the most effective factors that persuade people are not identical with the technically strongest reasons, because often one needs to see it before one gets it. - what are the likely pitfalls in the early days? - what are potential factors that might make D a bad choice in this scenario? I would like to use D certainly - but it is of course much more important that the client gets the best result, however it is done. - am I right in thinking C++ integration more or less works, except instantiating C++ templates from D? what are the gotchas? (I appreciate there is not so much to go on, and much depends on specific factors). But any quick thoughts and experiences would be very welcome.Just a thought, try getting them to use D for prototyping. Worse case they will play around with D before settling on e.g. c++. Best case scenario they'll move it into production.
Apr 14 2015
Rikki:Just a thought, try getting them to use D for prototyping. Worse case they will play around with D before settling on e.g. c++. Best case scenario they'll move it into production.Good point. Pick a small part of the project (especially something I 'prepared earlier') and get them to feel comfortable first.
Apr 14 2015
On 15/04/2015 2:17 a.m., D Denizen since a year wrote:Rikki:That too, but I was thinking more along the lines of writing some business logic. It is very possible outside of a cli app or even web that they would want to use c++ for the user interfacing. The very worse case scenario here is that they get to learn a new language + possibly write the business logic faster then normal. After all, business logic probably isn't already designed as pseudo code.Just a thought, try getting them to use D for prototyping. Worse case they will play around with D before settling on e.g. c++. Best case scenario they'll move it into production.Good point. Pick a small part of the project (especially something I 'prepared earlier') and get them to feel comfortable first.
Apr 14 2015
On Tuesday, 14 April 2015 at 14:21:26 UTC, Rikki Cattermole wrote:That too, but I was thinking more along the lines of writing some business logic. It is very possible outside of a cli app or even web that they would want to use c++ for the user interfacing. The very worse case scenario here is that they get to learn a new language + possibly write the business logic faster then normal. After all, business logic probably isn't already designed as pseudo code.Yes - the business logic is where I come in, so it makes sense from that perspective, too.
Apr 14 2015
On 15/04/2015 2:24 a.m., D Denizen since a year wrote:On Tuesday, 14 April 2015 at 14:21:26 UTC, Rikki Cattermole wrote:I'm available for hire if you need a developer/consultant btw.That too, but I was thinking more along the lines of writing some business logic. It is very possible outside of a cli app or even web that they would want to use c++ for the user interfacing. The very worse case scenario here is that they get to learn a new language + possibly write the business logic faster then normal. After all, business logic probably isn't already designed as pseudo code.Yes - the business logic is where I come in, so it makes sense from that perspective, too.
Apr 14 2015
On Tuesday, 14 April 2015 at 12:08:54 UTC, D Denizen since a year wrote:Hi. I have been here a year or so, and trust you will forgive my posting pseudonymously on this occasion. If you guess who it is, please be kind enough not to say for now. A friend has been invited to be a consultant for an investment bank that would like to build a set of analytics for fixed income products. The team is currently quite small - about 5 C++ developers - and the idea is to start with a proof of concept and then build on it as there is further buy-in from the business. Having been using D for a year or so, I am pretty comfortable that it can do the job, and likely much better than the C++ route for all the normal reasons. I haven't experience of using D in a proper enterprise environment, but I think this group might be open to trying D and that I might be at least part-time involved. I also have little experience in getting across the merits of this technology to people very used to C++, and so I have not yet built up a standard set of answers to the normal objections to 'buying' that will crop up in any situation of this sort. So I am interested in: - what are the things to emphasize in building the case for trying D? the most effective factors that persuade people are not identical with the technically strongest reasons, because often one needs to see it before one gets it. - what are the likely pitfalls in the early days? - what are potential factors that might make D a bad choice in this scenario? I would like to use D certainly - but it is of course much more important that the client gets the best result, however it is done. - am I right in thinking C++ integration more or less works, except instantiating C++ templates from D? what are the gotchas? (I appreciate there is not so much to go on, and much depends on specific factors). But any quick thoughts and experiences would be very welcome.A couple of big pluses: 1) Ease of changing code. D codebases tend to feel more flexible than C++ 2) Easy to transparently make use of highly optimised low-level code in high level constructs, whether that means carefully written D, inline asm or calling out to established C(++) libraries. Possible pitfalls: 1) What systems are being targeted? D on obscure big-iron is a very different prospect to D on a bunch of x86 linux servers. 2) Added maintenance due to language/library changes, however minor. Not a particularly big deal IMO, particularly when your writing a piece of software for in-house usage. 3) Limited support options. There aren't swathes of consultants available at any time to fix your urgent problems.
Apr 14 2015
John Colvin:A couple of big pluses: 1) Ease of changing code. D codebases tend to feel more flexible than C++Thank you - and yes, I agree.2) Easy to transparently make use of highly optimised low-level code in high level constructs, whether that means carefully written D, inline asm or calling out to established C(++) libraries.My guess is these guys aren't going to be writing much assembler, but everyone cares about performance at some point.Possible pitfalls: 1) What systems are being targeted? D on obscure big-iron is a very different prospect to D on a bunch of x86 linux servers.I will find out more soon, but I doubt it's old IBM mainframes (would guess linux, especially since it's largely a new project).2) Added maintenance due to language/library changes, however minor. Not a particularly big deal IMO, particularly when your writing a piece of software for in-house usage. 3) Limited support options. There aren't swathes of consultants available at any time to fix your urgent problems.Yes - I think the support/hiring question will be something of a factor. Let's see.
Apr 14 2015
"D Denizen since a year" wrote in message news:qpharcskwrbfgkuubfrw forum.dlang.org...- am I right in thinking C++ integration more or less works, except instantiating C++ templates from D? what are the gotchas?C++ integration can be made to work, as in data can be passed back and forth and most function calls will work. The gotchas are everywhere, and getting non-trivial C++ interop working is a fairly advanced task that will require changes on both the C++ and D side. Heavily templated code is especially difficult.
Apr 14 2015
On Tuesday, 14 April 2015 at 13:17:09 UTC, Daniel Murphy wrote:"D Denizen since a year" wrote in message news:qpharcskwrbfgkuubfrw forum.dlang.org...Is there anything anyone has on the gotchas for non-templated stuff? I am sure this would be of more general interest, since everybody likely has some legacy code they want to connect with. I get the impression the subset of C++ people often restrict themselves to tends not to favour heavy metaprogramming, but I don't know how much template stuff they actually tend to do at this place. (Will find out).- am I right in thinking C++ integration more or less works, except instantiating C++ templates from D? what are the gotchas?C++ integration can be made to work, as in data can be passed back and forth and most function calls will work. The gotchas are everywhere, and getting non-trivial C++ interop working is a fairly advanced task that will require changes on both the C++ and D side. Heavily templated code is especially difficult.
Apr 14 2015
- what are potential factors that might make D a bad choice in this scenario? I would like to use D certainly - but it is of course much more important that the client gets the best result, however it is done.You would have to asess if the analysis you should create is time critical. If it is, a possible pitfall is the GC, and you should try to avoid it from the start. Replacing the GC with custom memory management is a lot of work if you do it afterwards. If the analysis is not time critical, I don't see any problems with D. Kind Regards Benjamin Thaut
Apr 14 2015
On Tuesday, 14 April 2015 at 14:21:15 UTC, Benjamin Thaut wrote:My guess is it's not a major factor, but maybe can design from beginning possibility to use custom allocators. Ie the part where latency matters does not deal with monster data sizes, and if there is a part where data sizes are large then probably latency will be less critical. (Outside of HFT, big for this area is much smaller than what other people mean by big).- what are potential factors that might make D a bad choice in this scenario? I would like to use D certainly - but it is of course much more important that the client gets the best result, however it is done.You would have to asess if the analysis you should create is time critical. If it is, a possible pitfall is the GC, and you should try to avoid it from the start. Replacing the GC with custom memory management is a lot of work if you do it afterwards. If the analysis is not time critical, I don't see any problems with D. Kind Regards Benjamin Thaut
Apr 14 2015
On Tuesday, 14 April 2015 at 12:08:54 UTC, D Denizen since a year wrote:- what are the things to emphasize in building the case for trying D? the most effective factors that persuade people are not identical with the technically strongest reasons, because often one needs to see it before one gets it.From the business / management PoV I think it is important to emphasize low risk of D investment because of its relation with C/C++. This applies both to the fact that D code has strong interoperating capabilities with C/C++ (so any prototype effort won't be completely lost if discarded) and to the hiring process - it is possible to hire C/C++/Java developers and quickly get them into D development with a quick crash course because there are so many familiar things.- what are the likely pitfalls in the early days?Because total community / user base size is well below the critical mass anyone planning to use D in production needs to be read to invest into libraries / tools for domains that no one has been using before. I'd strongly suggest to do preliminary analysis of necessary dependencies and do estimates of what it will take to get it to desired in-house quality. It is very likely that overall productivity gain will outweight this drawback but it still needs to be taken into schedule.
Apr 14 2015
On Tue, 14 Apr 2015 12:08:53 +0000, D Denizen since a year wrote: I lead a group that uses D pretty much 100% for a number of different kinds of projects, so I've seen many sides to D's advantages. That said, I don't enough about investment banking analytics to know exactly what they value. Here are some suggestions: What are the sine qua nons of the project? * Are there hard performance requirements? * Are there specific hardware requirements? * Are there software/library requirements (e.g. you must use our proprietary ...) These are areas where you must prove that D is adequate, but not necessarily where you should focus. What characterizes 80% of the code? * Large linear algebra problems? * High-level business logic that needs to be maintained by an analyst? * Database operations? As an example, if the high-level business logic is the bulk of the codebase, demonstrate how D's cleaner syntax lowers the bar for code review by analysts. Show how UFCS and pipeline programming allow intuitive and highly-legible composition. Find some key pain points that D is particularly amenable to solving. If you can list a few of these sorts of points I might have more concrete suggestions. Cheers Justin
Apr 14 2015
I've spent a few years in the trenches so here are a few additional points that come to mind.. (+1 for John and Rikki's points BTW). -------------------------------------------------------------------- what are the things to emphasise in building the case for trying D? the most effective factors that persuade people are not identical with the technically strongest reasons, because often one needs to see it before one gets it.Extremely fast compiler makes for a very 'fluid' work-flow. The RDMD compiler wrapper turns D into very effective scripting tool Phobos standard library (subjectively) more useful for finance than Boost + C++ Painless C interop offers many integration possibilities Syntax is familiar curly brackets so will be familiar to C++/Java/.Net devs. Built-in slice syntax for arrays increasingly important, given that the array of contiguous memory is the new 'go-to' structure for cache-friendly algorithms. Paraphrasing Walter + Andrei's collective bios into 28 words or less can impress on management the credibility/gravitas of the language.- what are the likely pitfalls in the early days?Productivity will initially dip while everyone learns D, adapts to new tools etc., but will soon rebound to a higher level - so have expectations + timescales firmly in place + account for the initial 'dip' when making promises. Providing adequate tooling for a D-based group in an IB environment may be challenging - Linux distros can be antiquated/bad, unlikely you'll run as privileged user so a lot hinges on finding helpful sysadmin(s), desktop OS may be locked down. Entrenched IB C++ developers can be difficult to work with. They tend to cover a spectrum of personality types and abilities. Many are reluctant to change + are not particularly incentivized to embark on 'risky' projects. Many are simply intellectually incapable of learning a new language quickly. Best to find out quickly which variety you're dealing with + if they really buy into the idea. They can become the bane of your life if they don't.- what are potential factors that might make D a bad choice in this scenario? I would like to use D certainly - but it is of course much more important that the client gets the best result, however it is done.Any issues that occur will (fairly or not) tend to be blamed on having chosen a 'nonstandard' language, so there's an extra risk associated. But if things go well you can look like a innovator :-) Integrating with the IB's existing systems may be tricky. If you go 'off-piste' (so to speak), the onus will generally be on yourself to get things working; internal service/library providers may push back against supporting an unconventional language. When debugging a problem you may find yourself obliged to produce a minimal example in a 'supported' language in order to be taken seriously :-( Other than that I can't think of any pure technical reason why D wouldn't be suitable :-) ------------------------------------------------------------------- These are first few initial thoughts based on what you've said, think there's not much detail can be supplied without knowing a bit more about the specifics of what you'll be looking to do. HTH + Good luck! Cheers, A.
Apr 15 2015
On Tuesday, 14 April 2015 at 12:08:54 UTC, D Denizen since a year wrote:A friend has been invited to be a consultant for an investment bank that would like to build a set of analytics for fixed income products. The team is currently quite small - about 5 C++ developers - and the idea is to start with a proof of concept and then build on it as there is further buy-in from the business.One of the biggest issues I can think of would be code breakage. While we're the point where most compiler updates no longer break my code, if you expect to use a codebase from 2 years ago without having to update your code, you'll be disappointed.
Apr 15 2015
On Wednesday, 15 April 2015 at 22:32:53 UTC, Kapps wrote:On Tuesday, 14 April 2015 at 12:08:54 UTC, D Denizen since a year wrote:In the last 2 years each breaking change I've encountered has been to close holes where latent bugs may lurk ... and often were lurking! I welcome these breaking changes in production code. bye, loboA friend has been invited to be a consultant for an investment bank that would like to build a set of analytics for fixed income products. The team is currently quite small - about 5 C++ developers - and the idea is to start with a proof of concept and then build on it as there is further buy-in from the business.One of the biggest issues I can think of would be code breakage. While we're the point where most compiler updates no longer break my code, if you expect to use a codebase from 2 years ago without having to update your code, you'll be disappointed.
Apr 15 2015
On Wed, 15 Apr 2015 22:32:52 +0000, Kapps wrote:On Tuesday, 14 April 2015 at 12:08:54 UTC, D Denizen since a year wrote:not that anything is forcing someone to update the working compiler. i=20 know some big codebases that still using msvc6 for the exact reason that=20 "knowing bug is better than new bug".=A friend has been invited to be a consultant for an investment bank that would like to build a set of analytics for fixed income products.=20 The team is currently quite small - about 5 C++ developers - and the idea is to start with a proof of concept and then build on it as there is further buy-in from the business.=20 One of the biggest issues I can think of would be code breakage. While we're the point where most compiler updates no longer break my code, if you expect to use a codebase from 2 years ago without having to update your code, you'll be disappointed.
Apr 15 2015
On 2015-04-16 00:32, Kapps wrote:One of the biggest issues I can think of would be code breakage. While we're the point where most compiler updates no longer break my code, if you expect to use a codebase from 2 years ago without having to update your code, you'll be disappointed.I don't agree. Every single release since at least DMD 2.050 has broken DWT. -- /Jacob Carlborg
Apr 15 2015