Girba Pitches Moldable Toolkit to Slash Developer Code-Reading Time

吉尔巴推介可塑工具包，大幅缩减开发者读代码时间 由于开发者超过一半的时间都花在手动阅读代码以进行决策上，工程环境必须允许程序员仅用十行代码即可生成自定义可视化工具，而非费力解析原始文本。 都铎·吉尔巴（Tudor Gîrba）对软件工程传统上专注于构建的观点提出挑战，认为开发者超过一半的时间都花在阅读代码以做出决策上——他将这一活动称为评估。为了突破现代 IDE 的停滞状态，他通过 Glamorous Toolkit 引入了可塑开发（moldable development），使工具、对象检查器、搜索和文档能够动态适应人类认知和特定的领域上下文。吉尔巴现场演示了自定义视图、即席绘图、可执行文档和轻量级插件，展示了编写领域特定表示形式平均仅需十行代码。他总结道，由于我们的工具有形塑造着我们的思维方式，软件系统必须让工具具备可塑性，以消除代码、数据与人类理解之间的认知鸿沟。 开发者至少有 50% 的工作时间花在阅读代码而非编写代码上，这使得信息评估成为软件开发中耗时最多且成本最高的活动。 (0:05) Glamorous Toolkit 展示了可塑开发理念，它用反映人类心理模型的上下文相关、领域特定可视化视图，取代了统一通用的对象检查器。 (5:16) 创建自定义视图扩展所需的开销极小：在演示镜像中的 725 个活动视图方法中，每个自定义展现形式平均仅包含 10 行实际逻辑。 (32:34) 可塑工具通过促使开发者将算法和中间执行状态具体化为头等、可检查的对象，从而推动软件架构的改进。 (34:32)

[32:34] Tudor Gîrba: Tools are essential in software engineering and they should be moldable. [3:09] Tudor Gîrba: it's actually useful to start regarding software engineering as being, you know, less of a construction business and more of a decision-making business. [24:14] Tudor Gîrba: Once a human deems a representation of a fact as meaningful, it's the responsibility of the system to carry it forward. [21:43] Tudor Gîrba: Communication problems will come or are generated because of the differences in perspectives, but if we make those perspectives apparent, the conversation completely changes.

[0:05] if you have any questions he is welcome [0:08] to hear your questions during the talk [0:10] so you don't have to wait but because [0:12] it's recorded raise your hand and I'll [0:14] bring the microphone to you hello [0:18] everyone [0:22] one more time hello everyone and what's [0:26] better enjoying the conference excellent [0:29] so here's the thing today I want to sell [0:31] you on an idea and the idea that tools [0:36] are essentially software engineering and [0:37] that they should be moldable well I will [0:40] be demonstrating what that means but I [0:42] would like you to engage with me if [0:43] possible I want you to be acceptable as [0:46] you can and challenge me as hard as you [0:48] can is that ok ok good so my name is [0:53] tutor girl but I work at think that's a [0:56] company that I co-founded it's a [0:59] consulting company but our real goal is [1:01] to change the development experience [1:04] that is the experience that people have [1:06] when they work with software systems [1:09] notice how I don't necessarily say [1:11] developers because I think that [1:13] everybody that works with the software [1:15] system should be integrated there and so [1:17] when we talk about the experience we [1:19] talk about an IDE right and so this is a [1:21] typical kind of an IDE all right anybody [1:25] familiar with that good right so so this [1:29] is one of those and here's another one [1:32] and what's interesting about them is [1:34] that it's hard to this to distinguish [1:37] between them right if you squint from [1:39] afar it's it's not really easy they're [1:41] very much similar and here's maybe [1:43] another one so the question here is is [1:46] this the pinnacle of design and if not [1:49] why did we not evolve over the last 20 [1:52] years and so is it possible that may [1:55] perhaps we're asking always the same [1:58] question and that prevents us from [1:59] evolving so I'm gonna suggest that this [2:02] is true so again before we dive into the [2:06] actual new proposition [2:09] I would like to frame the problems like [2:10] the different thing so here's the thing [2:12] all of those tools they they are called [2:17] development environment I mean whenever [2:19] we talk about development we think about [2:22] the construction part the empowering [2:24] part that is I don't have it I develop [2:26] it and then I have it right it's very [2:29] empowering and this is amazing because [2:31] like we're developers and we're just [2:32] everything around us gets completely [2:35] reshaped because of us so that's an [2:39] amazing thing and an amazing thing to [2:41] think about but at the same time if you [2:43] look closely at what developers do you [2:46] know developers actually don't develop [2:48] most of the time that's an interesting [2:52] thing so for a lot of the time 50% or [2:55] more according to different studies and [2:57] from my conversations with about 3,000 [3:00] developers so far developers spend 50% [3:03] or more of their time reading code would [3:06] you agree with this right now here's the [3:09] here's the other thing but just could [3:12] you keep your hand up please if you [3:14] agree with this so if you do agree with [3:17] this right this is 50% or more of the [3:20] time this is the largest most expensive [3:21] activity we have so when was the last [3:24] time you heard developers talk about how [3:25] they read code right it's amazing right [3:30] very few are still here right even if [3:32] you did hear them keep your hand up if [3:35] you've heard of like a very regular [3:37] conversation or did it just happen last [3:39] year or half a year ago it's not a [3:41] subject of conversation right so few [3:43] hands are which is great but in general [3:46] it is not a subject conversation thank [3:49] you so that's was the this was your [3:51] sports for today so what is the problem [3:56] of this well again this is the single [3:58] largest expense we have and nobody talks [4:01] about it and I think we should now if [4:04] you do talk about it then in fact most [4:07] of the time they really you know the [4:09] reason developers read code is because [4:10] they want to understand what to do next [4:12] right they want to understand how do you [4:15] know what code to make a decision [4:17] it's called decision making so the [4:20] activity is actually not reading the [4:22] activity is decision making so perhaps [4:25] given that this takes most of our time [4:26] perhaps it's actually useful to start [4:29] regarding software engineering as being [4:31] you know less of a construction business [4:33] and more of a decision-making business [4:35] and so if reading is the strategy then [4:39] perhaps there are other strategies right [4:41] because it's a strength this would be [4:42] the strategy with which you extract data [4:45] or sorry information out of data because [4:47] that's what you need to make decisions [4:48] it so happens that reading is actually [4:51] the most manual possible way in which [4:53] you can do that right so then why are we [4:56] not constructing a different perspective [4:57] on this given again this is the largest [4:59] expense so the first thing we should do [5:01] every time we want to talk about [5:02] something is we have to give it a name [5:04] so I call it assessment and that's I [5:08] think that's a core activity in software [5:10] engineering and we should actually [5:11] explicitly talk about it and learn from [5:13] each other in doing that ok so that's [5:16] the perspective from which I want you I [5:19] want you to looking at the two demos [5:21] that I will be giving you so I'll be [5:23] showing you glamorous toolkit that's a [5:25] new kind of a development environment [5:27] which we call it's a moldable [5:28] development environment and I'll explain [5:30] what that is um [5:32] glamorous toolkit is an environment that [5:35] is built in thorough and at the moment [5:37] works for faro anybody knows what far [5:39] away's Wow how surprising mom so far Oh [5:42] for those that don't Farrow is a [5:45] beautiful object-oriented language is a [5:47] small talk system and it comes with its [5:50] own environment the language plus the [5:51] environment and what grammars toolkit is [5:53] is a replacement of the environment it's [5:55] a complete new graphical stack with [5:56] complete new kinds of properties inside [5:58] it enables new workflows so as I said I [6:03] don't want to focus so much on how we [6:04] construct systems and much more than [6:06] tall spend the time today to look at how [6:09] we reason about the existing systems and [6:11] in order to change bigger you know [6:13] bigger patterns if we I want to start [6:16] from the cell so given the first opal in [6:19] this case we are looking at an [6:20] object-oriented system but also in other [6:22] cases the smallest possible thing that's [6:24] most basic thing that we have is perhaps [6:27] not computation but data right and in [6:30] this case in an object-oriented system [6:31] it would be an object so if you look at [6:36] a typical IDE and how objects are [6:39] represented so when you open an [6:40] inspector for example you see something [6:42] like this [6:43] right so here I have six different [6:45] inspectors showing you six different [6:48] objects and when you see right they're [6:51] all representing the same way variable [6:52] and value right that's a that's a [6:55] reasonable way to look at this right but [6:58] it's only reasonable if if the thing [7:01] that we want be expected to be is to to [7:04] be something that shows accurately the [7:06] interior of that object but most of the [7:09] time when we want to work with an object [7:10] we don't necessarily care how it is [7:13] represented internally we care about [7:15] what it means for us it's a different [7:18] perspective so what if I take those [7:21] things that you know this representation [7:23] works equally well for all objects [7:25] that's why all inspectors in the world [7:26] basically show it the problem is that [7:29] most inspectors in the world today and [7:31] stop there but let's imagine I have [7:34] these objects and I change it like this [7:36] how does that make you feel did you feel [7:40] something here not seriously did you [7:42] feel something it's the same thing right [7:45] it's really literally exactly the same [7:47] object these two are the same objects do [7:52] you feel how your brain completely [7:54] changes how it engages with that those [7:56] things they are boring things otherwise [7:59] and immediately they change they change [8:01] their nature just because we change the [8:04] shape we molded the shape into something [8:07] that is more comfortable for your brain [8:09] and all of a sudden you could engage it [8:12] changes the way you engage with the same [8:14] problem so that's actually an [8:15] interesting thing because here these are [8:17] not these are not pictures these are [8:20] actually these are actually live [8:23] inspectors there and when they when we [8:25] look at them they look like this right [8:28] this is a really living system and [8:31] you're looking at an actual object in in [8:33] that living system so what you see here [8:37] on the left you have a fraction and then [8:38] we have the color and then there we have [8:40] a folder there we have a picture then we [8:43] have some sort of an algorithm showing [8:45] us the neighborhood [8:47] finding the next neighborhood circle and [8:50] then here we have a data structure of a [8:51] rope of the internal representation of a [8:54] piece of text [8:56] okay so completely different things [8:58] things should look different so that's [9:02] that's the that's the core and it's a [9:04] core property yes we have a question [9:05] here thank you very much for being so [9:07] skeptical I'm not skeptical at all it's [9:12] awesome I was just wondering about the [9:14] this string is that some sort of why is [9:19] that represented as a tree like that [9:21] that's that's a rope data structure okay [9:23] yeah so a rope data structure is one way [9:26] to represent text in a text editor and [9:28] in fact you can look at it like this [9:30] this is a string representation of it [9:32] and this is a textual represent this is [9:33] how the text actually looks like when [9:35] you render it and this is the data [9:37] structure behind that text when you look [9:38] at it very cool things different [9:40] perspectives on the same thing so so [9:43] let's do something with this right so [9:45] now here what I have here is something [9:46] we call a playground so I'll just show [9:51] you a kind of an exploratory thing so [9:53] let's say that I need to go and look [9:56] into my file system so I'm just gonna [9:59] start from my local folder so I'm [10:02] inspecting a folder so now what do I see [10:04] here instead of seeing this which is [10:06] again this is accurate it's a completely [10:09] an accurate description of what is [10:11] inside that object but this is useful so [10:16] so because I have this distinction right [10:19] there's the inspector what I showed you [10:21] here is if I show you an inspector that [10:23] have first of all different views every [10:24] object every individual object can be [10:27] perceived from different views depending [10:28] on the context from which you want to [10:30] look at it so that's one the second one [10:33] is and I'm showing you connected objects [10:35] so here is acute it something and I'm [10:38] showing you the resulting object to the [10:39] right the other thing there the third [10:41] operator here is that when I select [10:44] something over here and I dive into that [10:47] it's basically the same thing so here [10:50] navigations are clicking and executing [10:53] code they are the same they are [10:54] equivalent [10:55] right which basically means that there [10:58] is no necessarily a distinction when [11:00] when the user interface stops I can [11:01] continue programming in fact we look at [11:04] the IDE as [11:05] language to an IDE is not a set of [11:08] features but as a language a language [11:11] made of visual and interactive operators [11:13] that basically can extend the host [11:15] language so let's see what what else can [11:17] you do here so here I have it I have a [11:19] text file so obviously I can see a [11:23] preview of those contents and here for [11:26] example I have a Jason so I can look at [11:29] this Jason and just looking at the [11:31] contents of the file but because it's [11:32] Jason I can look at it like this and now [11:34] this is an actual editor here so I can [11:36] go and change it now remember I'm in an [11:38] inspector inspector was supposed to be [11:41] the smallest or most basic tool that you [11:43] have at your disposal [11:45] so now you see I already see here this [11:49] is a Jason and there's an array over [11:50] here I'm made of some elements and I [11:52] just don't know how many of those there [11:54] are so what could I possibly do I could [11:56] potentially go and scroll here right I [11:59] could do that but this would imply now [12:02] that I'm gonna read my way to finding [12:04] out this answer and that's a highly [12:06] unsophisticated is it a highly manual [12:09] way and we should go we should go away [12:11] from doing that so instead because it is [12:14] Jason Jason is not a text right it's [12:16] super highly structured data so why not [12:19] you know treat it like that so here for [12:22] example I open another playground it [12:24] just works within that context so for [12:26] those that are maybe not familiar with [12:27] this motor system it's like a ripple [12:29] that you apply that it's you know that [12:32] where self is these object that you have [12:34] there now because we already have a [12:37] jsonparser I can just type here I want [12:42] this parser and I say I want to apply [12:45] this parser on the string and now self [12:47] is the file and if I say contents it [12:51] will give me that those contents so now [12:53] I'm parsing that and I get an object so [12:55] now I know they're exactly nine items in [12:57] my array all right then I continue I can [13:00] continue announce navigating like I I've [13:03] done right so much even if my user [13:04] interface maybe was not exactly complete [13:06] I could still catch those things with a [13:09] little bit of a code and then I continue [13:11] navigating there is no there is just to [13:13] continue there's no boundary between you [13:16] know the perspective [13:16] even if the user and then of a developer [13:19] let's that's all the core of of a small [13:22] talk system but we're taking it much [13:24] farther because we have been now [13:26] extended with the language a visual [13:28] language for the IDE now of course [13:31] because we know that this is a typical [13:33] activity it is this Jason from string [13:36] applying it to contents we actually have [13:38] here a little icon right so I can [13:40] actually part click on that one it [13:42] already has that same ability right so I [13:46] learn as I learn what I want from my [13:49] environment the environment grows too so [13:53] this is one one thing too that we could [13:56] do but here's another one it seems like [13:57] here I have a couple of folders and I [14:00] just want to have a quick overview and [14:02] see what you know how deep those folders [14:04] are I don't know so one way to do that [14:07] would be I could just build a map here [14:08] so I could just say let me just I'll [14:12] just open an object on which I could [14:15] just draw something at the beginning I [14:17] don't draw anything so I just want to [14:19] look at the data structure or the folder [14:22] structure for example so I'm just saying [14:24] here I want the nodes for something from [14:28] this folder over here and I want to show [14:31] in that folder all folders so now in [14:36] execute that it's not all folders it's [14:39] all directories now when I execute that [14:42] I get this thing so there are not that [14:44] many actually but let's see how they are [14:46] organized and so I want to create some [14:49] edges now and I'll just connect each of [14:52] these ones from its parent so once I do [14:55] that I start to have some little lines [14:57] there so maybe I will arrange them [15:01] slightly differently now so now I have a [15:04] tree and then I can continue but what's [15:06] interesting about this is that this is [15:07] not a picture right so if I now click on [15:10] one of those I actually go to where I am [15:12] so the tool can evolve with me it is not [15:16] a dead end right so questions so far so [15:24] here's an XML right I see that I have an [15:27] XML over here if I go into it again [15:30] I have you know it's a small catalog of [15:33] books with authors and titles and let's [15:36] say that I'm quickly interested in just [15:38] finding out what are the author's did I [15:39] have in there so again what I do is I [15:42] look at the XML but that's not a piece [15:44] of text it's highly data structure [15:46] although you might not necessarily like [15:47] it it's still highly structured quickly [15:50] quick question how many don't love it [15:51] for now okay good so yeah but we can [15:58] still parse it right so I don't have to [16:01] like it but it's a data structure and I [16:03] want to treat it as if it's a data [16:04] structure so now I'm gonna go and I see [16:07] a tree in there I don't see a piece of [16:08] text so I can still navigate there but [16:10] it's still it's still clumsy so what I [16:12] would like to do is I would just like to [16:14] go and pick only the author elements [16:17] from this tree so what do I do if I have [16:20] an XML and I just want to find some tags [16:21] out of it but what I do [16:23] sorry exponent right so here obviously [16:27] because if we have an element [16:28] I could just say here self XPath and [16:33] maybe say slash author when I execute [16:37] that one I get 12 elements but that's [16:40] you know that's interesting I could do [16:43] these and I could maybe but the thing is [16:44] I can't make this one on a I cannot make [16:47] this one on the button I need to map it [16:49] because I need an input I need a string [16:51] right it's not just a unary execution I [16:54] need to pass in the parameter so in fact [16:56] what I would like to do is I would like [16:58] to extend these ones so maybe here I [17:01] should get a new place where I can put [17:03] that thing so what I will do now in [17:06] order to do exactly that I will now [17:08] switch to the meta tab so the medaka [17:11] gives me the source code it's a browser [17:13] for the object that I am looking at and [17:17] then if I go somewhere and let's say [17:19] that I'm looking for XPath something and [17:21] I think it's probably in node XPath [17:27] should be right right so somewhere here [17:32] there is this method which is called GT [17:35] expat 4 so this is actually I already [17:37] implemented it right so notice how I [17:39] have here this self ex bath whatever it [17:41] says it has a [17:42] and over here with a certain I can and I [17:45] want to apply this button on a text [17:47] editor so now what I want to do is I [17:49] want to say I want to now just and [17:51] annotate the object with this idea that [17:55] you now can present to me than you idiot [17:58] so now if I execute that one [18:00] the magic tab appeared over here right [18:04] it didn't exist and now it exists and [18:05] now it has it's an editor so I can now [18:08] write here slash author and when I run [18:12] this one I get exactly that so what just [18:14] happened here while I was working with [18:17] the tool I notice that I want to extend [18:20] the tool and I take a little few minutes [18:22] and then I extend that tool and I [18:25] continue because then the way we look [18:28] the way objects and everything around [18:31] our software system the way those look [18:34] they determined the way we're going to [18:38] think about them and that's the premise [18:41] of multiple development so now if you [18:43] take a step back here you know what kind [18:45] of a tool is this right is it the file [18:48] browser is it a visualization tool is it [18:52] an editor or is it a query tool right [18:56] it's it's all together there is a [18:58] question there but this idea what I [19:00] wanted to show here is that there are no [19:03] tools silos tools file those are just in [19:05] an arbitrary artifact that were produced [19:08] by you know but essentially by two [19:11] vendors and I think that's a wrong way [19:12] to look at this problem yes thanks how [19:16] do I get like sharing the change I just [19:19] made to the rest of my team oh I will [19:21] commit it to get and then next bill will [19:23] just construct me that you and can I [19:26] still see or reject parts of the comment [19:29] sorry this piece of code is to those [19:35] together with the rest of the source [19:36] code and if I have it and I can put it [19:40] together with the rest of my system or I [19:42] can put it in a different package or [19:44] whatever I want I committed to get and [19:46] then whatever I assemble [19:47] assemblé right doesn't make sense does [19:50] it answer the question okay cool [19:52] now the question so far okay so let's go [19:56] to something so far I just showed you [19:58] some technical things so I want to show [20:00] you a couple of case studies because [20:02] we're using these tools already in real [20:04] life situations in fact we've been [20:07] developing those for about four years [20:08] now and we're this is the second [20:11] generation of the glamorous toolkit and [20:12] we are already using multiple [20:13] development since about five years or so [20:16] so there's this extensive experience [20:19] that we have with this and we we went [20:21] through many case studies so this is the [20:23] most recent one poly poly is is a [20:28] company at which we want to change the [20:32] the way the end user an end user takes [20:35] control of its digital shadow so this [20:38] idea is that your data is now this [20:39] person throughout all of these services [20:42] and they you know they basically control [20:46] they control your data and you don't [20:48] really have access to it but it's [20:50] interesting that in in Europe right with [20:52] the GDP our GDP our forces but vendors [20:56] to actually provide you the data so so [21:00] here's actually we have this idea of a [21:02] pod where it basically lives on your [21:05] machine and in this case I'm showing you [21:08] an inspector over that pod and this part [21:10] basically here I have a few accounts so [21:12] it's e to the rib on LinkedIn and fin [21:15] comm that's the company account on [21:17] Twitter here but on Twitter and to our [21:19] group on Facebook so these are different [21:21] accounts and what's interesting about [21:23] them is that the source code that comes [21:26] so the data that you download you [21:28] actually have to go and request your [21:30] data you download the data and at some [21:32] point it basically looks like this right [21:34] so if I look at the Twitter and I could [21:36] see I see something like that and that's [21:37] I'm familiar with because it looks like [21:39] a Twitter feed but at the same time the [21:41] raw data the data that you actually [21:43] download looks like this [21:45] so maybe here's the contact information [21:49] or the different tweets so they actually [21:55] look in this way right so this is [21:58] actually a Jason wrapped in a Java [22:01] Script I don't know exactly why but [22:03] that's what you get right so again this [22:05] comply is compliant to the law but it is [22:08] not understandable for a human brain or [22:10] for a regular human brain so because it [22:14] is not understandable for us either [22:16] we translate it alright so again it's [22:18] the same thing but now you look at it [22:20] differently and so this ability to [22:23] introspect and go quickly and then [22:26] change the perspective is super [22:28] important is even especially in these [22:30] kinds of cases and development is filled [22:32] with these types of cases right but you [22:34] have to jump the metal level very [22:36] quickly and what's interesting about [22:38] this is that again this is a detailed [22:40] representation of the data and we need [22:43] this representation but at the same time [22:45] this is a user perspective on the data [22:47] something a normal user would understand [22:49] without any kind of knowledge about the [22:52] any technical details so what's [22:54] interesting is because why do we have to [22:56] have this this distance between you know [23:00] a developer who's also happens to be a [23:02] human and other humans that are not [23:05] necessarily developers why can we not [23:06] bring those together and what is not [23:09] this would this not address the problem [23:10] of communication because communication [23:14] problems will come or are generated [23:19] because of the differences in [23:20] perspectives but if we make those [23:22] perspectives apparent the conversation [23:24] completely changes so this is one of [23:28] those cases in which we can use this and [23:33] so where the exploration part is really [23:35] really critical and what's interesting [23:37] about this flow is that actually we're [23:39] spending the largest majority of the [23:41] time the development time in the [23:44] inspector and not in the code browser so [23:48] we will literally develop for the most [23:51] of our time not even in the debugger and [23:53] not in the static code browser but in [23:56] the inspector where the presence of self [23:58] when you always have a self so so that's [24:03] one but I was talking about integration [24:06] at the beginning and I said that [24:07] integration is not necessarily something [24:09] that is that should only include [24:11] developers so [24:14] here's a here's a different perspective [24:15] like we we thought how do we integrate [24:17] norm developers so long technical people [24:20] not just you know maybe users and so but [24:22] on technical people people that [24:23] understand business so one of the you [24:27] know things that are the mediums that [24:30] get exchanged or that basically bridge [24:33] the gap between technical non-technical [24:34] people are documents so here we have a [24:37] different kind of a document engine so I [24:39] want to explain to you how we implement [24:41] it some prices so you saw there's a [24:42] markup language here a markdown kind of [24:45] thing markup language it's not marked [24:47] down and this editor basically makes the [24:51] markup disappears as I as I work with it [24:53] so let's say that I want to take an [24:57] example and I want again I want to just [24:59] describe who this is a financial detail [25:00] about how do we apply discounts on to [25:04] prices so what kind of design did we put [25:06] on that so what I will do here I'll just [25:08] open a piece of code and I will say here [25:12] I want to have a hundred euro as price [25:15] and as I type notice how these buttons [25:17] appear right I didn't there was no edit [25:19] mode and when I run it I get an [25:21] inspector of that one so I have here a [25:23] concrete price so again this already [25:26] exists this whole code runs this is a [25:28] real IDE running there is a real system [25:31] behind it this is another one so so here [25:35] let's say I'm one of these counties one [25:37] by ten euros and if I execute that one [25:40] I'll get a price is counted by money so [25:42] this is basically like a decorator [25:43] pattern how many of you are familiar [25:45] with the decorator pattern okay so yeah [25:48] and so basically this is another another [25:51] another object that wraps the original [25:55] one and applies the discount right so [25:56] like a functional decomposition there so [25:59] let's apply another discount here so if [26:02] I say discounted by 10% [26:04] I get 81 euros exactly what I expected [26:11] so now let's imagine that I want to [26:13] communicate this because we think that [26:15] you know software like we saw in the [26:17] keynote from yesterday [26:18] software like the only truth the main [26:20] truth is in the code but the code shows [26:23] you fact it doesn't it doesn't tell you [26:26] seasons it doesn't tell you why it is [26:28] like that and why it is like that why [26:31] something where anything is the way we [26:34] share it as a you know through our [26:35] culture is through narratives you want [26:38] to tell stories about the kinds about [26:41] the things that are less easy to [26:43] understand so that's what we think the [26:46] idea should be should be a platform for [26:48] storytelling - so here I want to tell [26:51] you a story but now it's a little bit [26:52] difficult because no I want you to tell [26:54] the story of how this is actually [26:55] created so now of course I can go here [26:58] say discount percentage and it's applied [27:00] to 90 euros which is the discounted by [27:03] money which is 10 euro applied to the [27:06] hundred euro which is over here right so [27:08] I could go and navigate from that thing [27:10] which was just top top down I go I go up [27:13] left to right but now I lost in the [27:16] context right so it's a little to [27:19] complicate it so now let's compare this [27:21] one with this view over here alright so [27:25] this view now basically tells me it [27:28] gives me that same object right but now [27:31] I present it differently so I have an [27:32] object over here which is 81 euro is 10% [27:35] off of 90 euros which is 100 euros - 10 [27:38] years ok now again this view only exists [27:43] for this object this is handcrafted for [27:45] this specific object but the interesting [27:47] thing is that the the drawing here was [27:50] drawn by non-technical people on the [27:52] whiteboard and once they drew it on the [27:55] whiteboard it belongs not on the [27:57] whiteboard once a human deems a [27:59] representation of a fact as meaningful [28:01] it's the responsibility of the system to [28:04] carry it forward ok so yeah we can go on [28:09] like this but so another another thing [28:14] here actually is that I don't want to [28:16] write any code into my into my [28:19] documentation because then I have to [28:21] deal with this idea of how do I [28:22] synchronize the code between the real [28:23] code and documentation and so there's a [28:25] whole set of problems but now actually [28:27] wrote this code so that was an example [28:30] that you saw there and now we are also [28:31] using these examples and so because I [28:35] it's like these are like tests but [28:37] they'd return an object [28:39] so here what I want to say is I want to [28:42] just sum on that example and I wanted to [28:43] come over here and when I type and [28:46] notice how there's no edit mode as I [28:48] type here the the the the code is [28:52] someone and comes there and so now I can [28:54] execute that code and I can get it there [28:56] right so I transform my live document [28:59] becomes live as I as I try to throw it [29:02] so maybe I want to say that I don't want [29:03] code at all or and maybe I just want to [29:06] say that I just want to have a preview [29:10] just to show me one of the tabs so I [29:13] just want to show the overview tab and [29:16] as I move away I'm left with the live [29:18] document right it's just a document that [29:21] has no technical detail I can just ship [29:23] to non-technical people and what is [29:25] interesting about this is the cost of [29:26] this document on top of after that if [29:29] the normal development is very very [29:31] small right so so here's an actually you [29:36] know the this one would look like for [29:38] the actual domain so this is these are [29:40] the prices this is an actual system that [29:42] we implemented these are prices that I [29:45] showed you and I'll just show you [29:46] briefly the system also has a workflow [29:49] system that you have in there so here [29:51] for example we have one of these it's [29:55] you know it's a cozy corner over here [29:57] with a kitchen and a guest room and some [30:01] some actors and basically these we model [30:03] them as being actors and so then we want [30:05] to see how what happens when somebody [30:07] places in order so in this case a user [30:10] Ellis gets placed an order over here and [30:13] then he propagated all the way to the [30:15] cook and so I mean I'm still in a in a [30:17] document but here there is a to-do item [30:19] and if I click on that to-do item what [30:22] happens is that the message propagates [30:25] further right so I can learn [30:27] it's an again it's a live documentation [30:29] is a real documentation of a real [30:30] running system but done in a with the [30:35] with the live documentation and the [30:36] interesting thing is that the same idea [30:38] of mouldability can be applied to other [30:39] things such as search so here for [30:41] example we if I now first if I look for [30:43] a class which is object I can see it [30:46] here I can or maybe I look for other [30:48] things as I can see multiple of those so [30:51] that's an interesting thing [30:52] let's say that I'm interested to search [30:54] for du which is a method that for for [30:58] traversing so maybe I just want to look [31:02] for the implementers of that so I can [31:04] see how many there are but maybe I don't [31:07] want to look for doing general I want to [31:10] look inside the collection because I [31:11] know that something is there so now if I [31:13] dive if I if I dive into the collection [31:16] now the scope of the search changed and [31:21] so now when I look for do it will only [31:23] give me the the it will only give me the [31:25] methods it will only give me the methods [31:30] from the collection class right so the [31:34] idea here is that not only does an [31:36] object know how to present itself in [31:37] many ways it also knows how to search [31:39] it's hard to get search so if I apply [31:41] this one to a domain so here I get my [31:43] pot this thing that we implementing for [31:45] the different accounts right so if I [31:47] look for the toolkit I can see the [31:51] different kinds of post this comes from [31:52] gear bar and this comes from a different [31:54] account it comes from Twitter so what's [31:56] interesting here is a now I am searching [31:58] through multiple of those all right and [32:00] again in the same interface but I can [32:02] equally I can equally maybe just go back [32:06] and I just say I will just go inside [32:08] this one and I will if I now search for [32:10] Jie toolkit I don't only get the the [32:14] searches from that account right so the [32:16] search is also scoped and it's very [32:18] simple it's what is interesting here is [32:20] that these are just some examples right [32:22] they didn't handpick I mean I [32:23] hand-picked those but out of many so the [32:26] idea is that if I take a string for [32:30] example and I inspect that one and I see [32:32] a presentation over here this [32:34] presentation and I'll click on to see [32:36] the implementation of that view right [32:38] this is the method that I created before [32:40] to to implement this view over here so [32:45] this method isn't is annotated with GT [32:48] view right so let's quickly find out how [32:50] many of those do we have so if I say GT [32:52] view over here and I'm looking for the [32:54] pragmas which is the name of the [32:56] annotations I can search for that and as [32:59] I search I should get a browser in which [33:01] I can go and navigate so this is an [33:03] actual editor in there [33:04] but let's quickly find out how many [33:05] there are and so they are 725 right now [33:08] right [33:09] so don't quickly find out how many you [33:11] know how many lines of code they all [33:13] have and I have here eight thousand [33:15] seventy eight thousand seven hundred so [33:18] if I do an average now so I get this GT [33:22] view and the GT proud mass and the size [33:25] here what I get is twelve twelve lines [33:29] of code of which two is one is a [33:30] signature and the other one is an [33:31] annotation so ten lines of real code [33:34] that's the cost of a little plugin so I [33:37] have 700 different plugins running at [33:39] the same time in this IDE in the india [33:42] in the daemon I'm giving you so the idea [33:46] here is that every time we talk about [33:47] tools I think we have to talk about [33:48] tools and every time we talk about tools [33:50] we also have to understand what Marshall [33:52] McLuhan and and his students warned us [33:58] from the last previous century and that [33:59] idea that was summarized by John Culkin [34:02] that we shape our tools and thereafter [34:04] our tools shape us right and what [34:08] follows there is that we have to be very [34:10] careful with the characteristics of the [34:12] tools that we expose ourselves to [34:14] because they will determine the way [34:16] we're going to think tools are essential [34:19] in software engineering and they should [34:21] be multiple thank you very much [34:26] [Applause] [34:32] we have a few minutes for more questions [34:35] if you have money everyone here I was [34:41] wondering impressive technology and [34:45] interaction [34:46] I was wondering how if one could isolate [34:52] this interaction and all it comes that [34:55] you said from the underlying language [34:57] yeah I think it's a complicated problem [35:00] but it would be nice if I could write [35:03] multiple languages in within getting the [35:05] same kind of feeling right yeah so [35:08] glamorous toolkit is actually a platform [35:10] I'd be very happy to show you that but [35:12] there's there's much more than meets the [35:14] eye there's actually already a software [35:17] analysis that form in there that deals [35:18] with us already many different languages [35:20] there is also extensive parsing [35:23] abilities and we will definitely target [35:25] other languages yes so this looks good [35:36] for visualizing different kinds of data [35:38] but often when I'm reading code I'm not [35:40] let's say you're trying to figure out [35:41] what's the data I'm working with but I'm [35:43] trying to understand how is the [35:44] algorithm work and what's the process [35:46] I'm going through this glamorous toolkit [35:48] help with that portion of trying to [35:49] understand the oh yeah so the way we [35:51] work with it and again grammar silk it [35:53] is the tool but what's interesting is [35:55] moldable development which is every time [35:57] you replace manual labor with some [35:59] automation the real benefits come from [36:02] when you when you replace the way you're [36:04] fake and so the way we work the way we [36:07] reason and build these kinds of views [36:08] you saw there are several twinning for [36:10] something views right so those 720 [36:13] something views in if you download the [36:15] glamorous toolkit you have about 400 [36:16] something which means that about 250 of [36:19] those are system specific so the [36:23] interesting thing here is that when we [36:25] work and we have a problem we don't just [36:27] solve the problem we don't just write a [36:29] test and solve the problem we build the [36:31] tools write the tests and solve the [36:32] problem and so now where is the problem [36:35] so if I have an object that is now [36:37] telling me or encapsulating an algorithm [36:40] and [36:40] becomes very interesting because there [36:41] is a there is a there's a very [36:44] interesting dialogue between the tool [36:46] because I sometimes want to explain [36:47] something but when it's not first class [36:49] it's very difficult to explain so then I [36:52] very often find myself having to go into [36:54] my algorithm make it first class often [36:56] maybe approach it through an account of [36:59] functional decomposition even if I'm in [37:01] an object-oriented language I work with [37:04] I work with we work with a lot with with [37:06] first class objects and then you [37:08] represent a step in that algorithm like [37:11] that or so for example I show you this [37:13] workflow right that was a one message so [37:16] we refined a message just because we [37:18] wanted to show you how the message [37:20] propagates through the different actors [37:22] and that was an algorithm so this is how [37:24] it works thank you very much again [37:35] you [37:45] you