Merken

Django: A Data Shovel With a Future

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
and her her time time time frame is time time
good morning everyone was my 1st visit that literally really happy be of didn't know already the city and and some of the 11 at all I will be going the big pico and conflicts in 20 16 17 the embryo to for
amendment I think about software along 2 axes of would be very conventional and just use the horizontal axis for time and I will graph vertically the idea of complexity having record having more complicated code it ended in general and we start writing code of a day month or a year of project looks something like this but over time becomes more complex sometimes the bumps in that trajectory but that's generally the shape of things and as I tried it a you get up in the morning I write a test or whatever I start trying to implement a feature I want and I write and I write new code until I reach this magical state called working code the and 1 of was a younger programmer I would stop there declare victory and move on it turns out that this was a mistake wants my code is finally working it is probably better
described as barely working and what have I done I have finally crested that the great water uh in the state space of of programs but from all the states my code could be even without having this feature I've just added on 1 of the states in which it now possesses the feature unless it's the kind of feature and very familiar with what are the chances that I will have crossed that magical states states state space but rather yeah at the best include list and most readable way to implement this feature the case of a very very small chance again unless I'm kind of already familiar and done the
trick before my guideline for implementing letter to me really new features was that once encode works but probably about halfway done with maybe I should have put it aside and read it tomorrow to realize how difficult to understand this code I've just written this but if have and an hour putting together what is to me a novel solution I generally find that in our later in the day year tomorrow really helps speed up and make it better hold so I
tend to start of course observer complexity clever Blake screener procedure in front of me I work and produce working code and that no image is a magnificent 1 because I can check and I compare plant a flag thank you plan to stay can say I at least now have this feature working by jury we find that the code itself needs more time it needs more work and that the complexity can then be ratcheted down through refactoring precaution through on clean up and that it's only then that I typically reach what I would call a good code web
framework history I think follows a similar pattern but built in the industry at large and also if you narrow your attention to Python itself are we started with the idea of of course in the very early days you just served the web of of your file system includes scripts into that file system if you wanted to random but once we got the idea of putting part software not the final system in charge we started writing code and we wrote more code and more coding magic to produce forums and magic to persist objects and magic to let you know as travel across our object of persistence hierarchy and we had finally but that mountain of cold all the way to the sky we give the name and the square and and it worked and what you intend we had we had reached this point where we could now go out and accomplish things on the web but just like when I'm sitting and in writing of a new feature for a few hours we didn't necessarily land on the simplest answer to the Web 1st and are blunts Ruby on Rails has inspired a community with this idea that the Web could be radically simpler than we made it our right out of the gate just a few months later but rather rural I believe came out with cherry pie which still hasn't following today and and that gender very quickly followed that it's interesting that there while other attempts at web framework since then but 2 of the ones left standing or those 2 1st responses that our community need to read a put you apply a little higher on the complexity scale because it's pattern was that you build or objects in New you kind of surface down the attributes of the control of objects with its path components Jacob kind of what I like steampunk and just you compare agreed well as with regular expressions you treat them as strings and that is a simplification or having to teaching programmers how bright objects course the communities experience didn't stop there just as I genderlect it's open said I think the red is simpler than that of course the micro frameworks and looked at soap shook their heads and said I think the red is simpler than that and produce faster produced by and the other are small and what cheek of micro frameworks that exist or should for the sake of records say that I have at times used a micro framework I generally find that if I really need to add a new line in a templating language and a cross-site scripting vulnerability package that by the time of kind of put that and kind of back up that changes complexity love at that resonant reach that complexity level gradually on demanded from the beginning of something instead in my system of server to and with are actually generally find the microphone works on more valuable because they let me pick the pieces that put together not because you're aware that is going to be noticeably simpler when it's finished in a micro framework if you used Jane Doe and finally I should mention a few programmers did stagger alive out of the collapse of the chain of others so and that of its code and they have actually are themselves not itself part about how they can solve those seen very CMS very content centered problems not simply and actually another agenda of other small world frameworks have died as that an the event was everyone I know in that community has now gone using pure that it's worth looking at if you want to see all the different approach to what you're you Wales mean so that is
kind of the trajectory I think that we've got taken gender was a 1st very successful experiment in how the web could maybe be simple now then a web framework but is this task that we can do in a complicated way for a simpler way I think of a web framework this way they web framework is a denormalization engines it gets under hopefully orthogonal independent and unique in our back in storage and it matches it and mixes it up into a single placed into a document that is delivered to consumers to the user and read the framework deal normalizes data so that lands in documents but which of course was on the member the prior problems of gene normalization as soon as my username doesn't just exist in the database but on a webpage that I haven't been in front of me I now have a copy of my data that raises all of the normal problems of caching what was there that will keep that username in the corner of my screen equal to the value in the database if it starts changing and so you have the further interruptions by far and away from as many as the tabs open it's just going to be out of date there is a way to use and usually that's where they promise automatic updating the data on the page when it changes on the back in the 1st spectrum on the art solutions to having a cash or having a copy of their of data a rise and the full range of simple too complicated solutions to that are inherently present in other web framework does which is made copies of your data usually being when analyzing orthogonal data on the back and uh into some more complicated for it and I've found especially as mobile devices have gotten better band with his improved storage has increased that the question the real uh increasingly commands my attention is the question of how much data in a particular case should you delivered to the client you can give a client a web page the absolute minimum just the exact date and no matter what the user has asked for and other and you can just do the entire database to java script and what the front and deal with it if you've ever use the Sphinx documentation system to burn the static site and notice that it has a search for you don't even know its static if you look behind that during the document building process the birds and index of the words in your document as the juice and Robert and stores it to a file that's search feature does not go talk to a search engine that java script reads in your entire the text search database as adjacent blobs and does the search term action right there in the front end so evil and on a Python project you might span in the solutions you're using the this distance spectrum between very carefully designed web pages that just give exactly what the user is asked to see that neurons and other parts of your solution but he essentially that a base on to the front end and let it do its own search imagine for a moment that
success was an respond think of what we've accomplished and what the future might be like in this industry what if hard drives kept getting I am told that they shall not do so because of things like physics but it's not that you know the we store data with magnetic polarity is I'm talking the starting to run into density problems at like you know the the atomic level and also we we can build an and purify platters and so forth about the researchers looking completely radical other ways to store data which we will be reached is enough it stated instructed the next generation of the drives store data on an atom-by-atom basis obviously it will take a long time and you wouldn't hold the can't expect the rate of improvement to keep going because we've essentially been writing a single solution for 30 years and it was really slow start this we have the it to something denser than magnetic platter mediums it just for fun modified rice kept growing the last 30 years from 1984 to 24 2014
hard drives have gone from the biggest 3 and a half inch of her drive being 20 megabytes to Seagate week 2 ago announced and 8 terrabytes derive that's been no Western Digital least for a few weeks after those clients that need maximum storage or the next 30 years what if they were able to keep going how big is the ratio between 20 megabytes and 8 terrabytes so if you do the math and we'll find that if we did this all over again but over the next 30 years about the time of my retirement idea but to by an off the shelf 3 . 2 x Exabyte however right I could put Facebook on a have time it's good so we have these and I estimates vary but some people think the NSA would need as many as 20 or 30 in order to operate and if that happened in if I live to see the day at least 70 years old a picture and whom to move my new through the Exabyte drive at the kids bicycling down the street and set a and then we used to have a maximum of a computer's despair this much data and the children will laugh at me that I because the young and because I will be but for the record similar and to look I meant to look like a good futurist here if this actually happens for the record I don't think that a 3 . 2 x a by device would be up your storage I think we'll probably go ahead and sacrifice some of that storage to put an array of a few thousands microprocessors on it so that instead of a single out that's trying to redevelop the zettabytes out to do something to reduce the data uh I believe you'd probably have small 1 3rd of processes a cluster so that every few petabytes could have its own dedicated processor what to do with that but even with the rate slows up we have a lot of storage potentially in our
future then we can today questions begin to come up as I sit waiting for a page talk they don't I have StackOverflow on my hard drive with the baby the differently maybe that marriage that by a given node and couldn't handle it in there overnight bag why isn't it the Wikipedia on my hard drive the rise of the MDP which you would actually is not that big it's a few other they had it is text files and you can download it and and process the data yourself like I sit and wait for any of those banner ads to load when I just want to see that guy's name who was in that movie with that girl the next
feature well wonder why not put the Library of Congress's media holdings video and then add here and of books on hope there'll be questions like I went and looked at the i in DB database added up the number of minutes of footage you have the movies and television shows in history you could have it all in high definition in it by my calculation about point 0 8 exabytes maybe maybe many of the selling points for the drives is that just come with an and movies ever petition and on the number the UK after the probably never notice if you are not buying it it for consumer purposes anyway on the news updates as new movies are made I know Hollywood would continue making movies over the next 30 years maybe it'll be point 0 2 x abides by that yeah you might think I'm
being silly but we have also ready made taken this step that and made this choice without version control you think that the latency of browsing between StackOverflow pages doesn't add up to much but do you remember what it felt like when you were never again within a few seconds for subversion to come up with an answer for you do you remember what it felt like actually that's not you remember using that for a moment that novel that quite manner that you remember is going back to subversion for the 1st time we use the data came when you simply had wikipedia i in Debian StackOverflow annual have drive I believe that when you went back to mobile device and browse the next day you would have that that's called the subversion feeling that feeling that your whole life is once again spent waiting so in 1 very important area how we manage our projects we've already made this flip we've already made this and that from the idea that I an application should just give me what I need at the moment to how about you just give me your entire database yes I'm not likely to look at the 1st 100 commits to Django how I you give them to me anyway just give me the whole history and it down on the client I will deal with the question of how much of that I look at incident does occur to me that there very nice privacy consequences to just having the Wikipedia on my hard drive and only my processes knowing what if it I'm looking a big room of
Django as I have used it in projects is the Django is flexible it is not opinionated about where my views land on the spectrum between share minimum and share everything but now it's true that there are helpers Django has class-based generic views that are opinionated about the relationship between EU and what kind of the database and trying to get to from it but those are the exception and Django there's something popped into when they make sense unlike some of its predecessors gender the framework itself is not strongly opinionated about letting you know that means I am beginner function definition and Ireland invited find routine to invent from the ground up what information lived there and what it will be we tell a
story about jingle project I did recently what we're of a very modest and small version of this paper well not radically simplifying the application was
a project for the New England Wild Flower society but they remained that in 1900 when they were founded in that there's a conservationist urbanization 1st came to be but that since then other known by the name of that actually do all kinds of conservation study of native plants study of the way that invasive species in garden species threatened New England's native of BIOMS and all of the other animals other creatures that depend on them and they have a big audience of students that use the resources amateur botanists and teachers who wanted to serve better than went and
found on the Boston but based on a Python consultancy up uh group that started back in those a heady days observed open Paul and those are the big solutions but has recently been using all gender 1 quite a number of the projects and having very great success with it and uh uh of need on E. I. to me that there is a developers were busy here had several working on this project but needed 1 more and so I got to then I got to help with this really really fun project it tries to solve the following problem botany and anim been particularly of botanical identification is traditionally done with work on the dichotomies keys on a dichotomy as actually it describes a a plant or something like that but every time it branches branches not several ways by exactly 2 ways the Dion dichotomies tells you that your own ways splitting into 2 choices over and over again before you reach limbs a twigs are at at the at the end of the process and dichotomies keys that kind
of a Choose Your Own Adventure series forever botanists they start at a higher level like plants end of the buttons designing the asks what are the 2 biggest groups uh in the narrative of plant categories and develop the question but if you answered correctly and I'm going to answer correctly do you look at the correct half if the plant kingdom unknown whose moments of species your plant wise the next question tries to split into 2 big pieces and then again and again that actually don't put the dichotomy scheme of New England native plants on their website here is 1 example of this is
the top level for gymnosperms and things like pine trees and key NGOs that a fairly primitive limit the seeds growing right out visible in the open and you've got to do when you that the gymnosperms is start by determining whether and playing you're examining seed Singley partially concealed by a red fleshy and whether the abaxial surface of the flat leaves his bearing pale yellow longitudinal stromatolites you should see middle schoolers faces light up when they're presented with this kind of question this helps me expert they look at that and immediately picture of that that group of plants and then know which way to go and he reached this particular page of the Choose Your Own Adventure this is not however something that tends to help students learn wildflower society when needed to address this
they wanted to focus on something that would allow and support in what the user directed search search but does not but descended step by step into the classifications biologists find useful but the kind of search the can't the project good botany that would simply ask you to choose a filter and that would let you look through a menu and say well I like to think I have in front of me has red flowers and that would narrow the set of New England species to just those when you could then say OK has smoother leaves and suddenly have you work result set at narrowed down more considerably I'm on a trip to Vermont and this is what I'm seeing the plant and that can narrow the range of species they wanted
the users to have on the left-hand side of the screen a set of very common questions and many get what questions button that lets you even ask you to rise in that movie I haven't really a belief in front of me given what we've questions give more system questions interestingly enough and told that the early user tests proved that 0 per cent of users ever notice the get no questions but I don't know if the effects that out but as you can see when you select 1 of these I filters it gives you the multiple choice questions and when you select 1 of the options and hit applied
are but his age of adjusts our in wanted to show only the species that match In fact if there are few of species on the page to make this possible the actual animates the format off the screen the ones that didn't fit and the tightening up of your results at the ones that do the I did that I it's a java script
if In the 1st iteration we did something that seems so obvious in retrospect and around the 1 of those same optical that moves it was 1 in my case it was 1 of my 1st times writing web API and we thought that to build a weight where readily PI you back with this kind of obvious one-to-one correspondence between the action the user is going to be performed by performing a and you have performs an analogous API call we thought that you could just study and note that list of the things that the user would be doing and just build a URL query that we do each of those and that that would constitute the design the let us explore whether that is the case if the person selects red flowers smooth leaves the plant that's present in Vermont thought right was there a query numerical where you simply concatenate 1 after another and all the questions that the user has answered so far about the plant and submit that whole question back to the server for evaluation to get the list list of plant ideas that should be present on the screen to we noticed that during a three-way query against a database but by mentioning it was at the through a joint with the same table appearing 3 times and then of course the student chooses a FIR filter and arguing a for join 5 way join six-way join and that's a big group of plants you get up to 10 way joins 12 joints but a that the searches can a little bit expensive and so we started wondering well what if we put some caching headers and maybe have I'm sure CDN a tentative content something at work in front of our ap and as a delivery of content delivery network thank you well that can showing helped reduce the server load about getting to our I keep parroting back a particular search resulted to students at 3 students offer students are the same person market began to trying to identify a plant happens to reach the same state in the search process answer no and the
reason is that so for 1st answer that was easy to address is that the exact same search depending on the that they choose these filters that can come out and in fact in factorial different ways that can be
solved through a careful practice called canonicalization where instead of allowing your application to have several different names for the same thing you instead insist that or if the user creates a query if 2 users create queries with the same meaning necessarily that produce the same text at the same query uh very
uh trick with this kind of you know was to restrict your script so that you always whether those filters alphabetically as you assume that you now which ever greater those clicked on these and you get exactly the same and have the possibility of holding the cached version of that result
but in our case and we probably only would have needed to do that had begun down this road in the case of our front in Java Script we're kind of on the client the of those that if letting the proteins build and deliver these you as how can you make sure that they are killing the memorized cached version of the URL and the other in factorial ways it could be written up 1 possibility that would have been on just in case you're ever going this direction that you could redirect the other you our hours so that on the people who failed to properly canonicalize that list them that quickly uh moved over to the you where you might have cashed with a result
on of the interesting things about this transform from we noted is that had we implemented that is purely textual and does not need the database the fact that that query at the top should really be that query at the bottom the actually has no bearing on whether this is a and actually you where effective for this is a transfer 1 a purely based on the fact that we dis clarify the user by looking at it this is the kind of logic that could be pushed out are you into the front end up catching and minimisation layer and not even a hit the server that's been answering canonicalized URL models with content on the 1 of the interesting things is might think that that all have an app sitting there answering through your lungs out they 3 islands if you read the standards can also be cached dimension of certain spit back 3 learned barely if you let it catches a solution like vastly would do the same about tension on 200 results can be a big deal and I fully we often a busy setting where the right cast shadows on a 200 results and I'm even thinking about it and of the others in the really really great red thread on the wounded several years ago when they switched off of Drupal content the only uses gender and why it matters to us with this is a free fair of the answering all the skeptical web developers are an answering them about why this was a really really good at moves that they made you know the most interesting part of paragraphs to come out of that
because the answer is to let the biggest performance difference around at making which actually actually have to do with Jenova's specifically they said the biggest performance boost of all was catching fish of course yeah singling cache-control headers to the content delivery network and therefore they often what's right of you you really you might very rarely think to customize the winter Django customize their fearful that a custom of view and they actually
started Tashkent don't have time that content delivery network and then don't come back to this every time for this year we give you permission to remember from a while so that we don't have this particular matter and the unknown having gone through so many different uh but room versions of the website since the late nineties had a lot of you know there's that had just then mistyped into popular web pages as where you has that made no sense you as the very 1st generations of web spiders had accidently put together by misunderstanding harder you know as we work and that we're just still out there are often a site that was uh and they had a lot of links the spiders within every few minutes are that don't exist and the never exist in it can be deciphered can be redirected to real content anymore and they found by giving the city permission to cash that the outgoing bandwidth was reduced by 66 per cent concerned about it must have good cash on your to hundreds on your web server might not have much left to do once the 220 and onions web pages are out there in the CDN being delivered from the relocation geographically close to the customers they found that most about the server was that it was 1 of course and uh the uh drop the lower average by 50 per cent by reading and non pages be cached by the CD and so if we
had wanted what we could have some that improve the state space even for third-party clients by reading these normalizations the cash yeah
but even if we do canonicalization and catch the pages behind these you ourselves How often would they be revisited in the answer turns out to be fairly where warily what are the chances that 2 students well in their careers landed exactly this query especially as you get to 5 and 7 and 9 different features of the plant having been entered imagine
a group of plants that have a hundred different filters you can choose in use by students who maybe only a cat I 5 filters before the head new plants enough Elizabeth underestimates but they had already given
number 100 choose 5 for the number of canonicalized problem canonicalized URL as those students could hit as the rendering out into the state space of this just going but the playing with biological classification that 75 again and not only a 3rd of 5 that searches deep students often go deeper in the state those of 75 million ensure there's a lot of folks it's very obvious to note that the shape and so are going to be some of these years has a hit more often than other others but you lined up with a very very big state space that your users are wandering through the year level rather than a lot of different information to catch so the amendment came when we conceptually kind of step back I think it was actually doing 1 of those freewheeling 20 minute Skype sessions this week and was intended to be a standard meeting that somehow often became a little bit more and 1 of the developers ask a question but caused us uh to think this way this year and this asking out this asking Django is optimally asking the sequel database to different way join against the plant versus feature binary relation at the bottom it looks something like this name the 3 that's in Table 3 times make sure that you're using plant ID to build triplets of revenues from those 3 tables and uh then limit each of those images of the table to family containing rose for 1 of the filters that you're looking at and it's that bad that's doing this work for your user it's looking at
the table of a plant species and their associated features this binary relation you use using a
where class to grab what say or the species that are trees rather than something else and then it is during the
joint is doing the intersection to look for the planned ideas in common between those 2 those for those 5 different on of virtual over the computed tables in your select statement and the developer suggested the following pivot instead of doing that so instead of doing what we had planned having the database of equal sign there mean during a straight comparison on that database table to get some lists of matching plants then during the intersection between those lists of plants to get only the plants that fit a 5 of your filters and having generated that answer handed back to the client what if we do this nonsense exactly the same thing at exactly the same computation but what if we just draw that
line some what if we instead of have the server stop with having generated a list of plants that are trees with the list of plants with red flowers given those to the client and Java script do the dreaded joining the java script take that step that actually produces this on very very broad state-space choose your own adventure state space that the user is
exploring we found that we actually have had a Jason payload that for a particular feature is this is this is a really good that will pop up menu and choose the leaf shape this sort of adjacent look like behind it with the person clicks and the shape of the JavaScript those and hits so the the edge you and previously not back some text and maybe a little image with which to illustrate those options will now just went into that Jason and we went ahead and we give away the database we just went ahead and for each of those options we went ahead and just the client and the of the select some really this was and the species in New England that that have that feature if you select a jagged believes that come and ask us we're telling you up front here is the complete list of species that match that filter a
single line of code if you use underscore JS will get those lists of numbers and fairly efficiently mix them down even kind of works on mobile to artists in the list of species it's not uh it goes through a list of integers and looking for similarity is actually something that computers are good at and that we found you don't even know most of the time it takes a job script to do this quick operation and 7 are search URL was gone this disappeared from the project and instead of the user in any water went through the color of a flower the shape of a leaf the state that they had found it in korean not I mean not like yeah but affective states if they went through these in any order exactly the same 3 as would be poured once and they would contain exactly the same information whether there were fetched very early they're very rate in that particular students search process suddenly the bad we were delivering was independent of whether the user was in their search and at the
front end could put together the list of 55 matching species by doing its own work to combine those possibilities that have been given
free features we have in your house end of story you know exponential space to have to catch and actually kind of noticed at the end that we didn't need gender we could actually generated generated as a static site because except in the biologist update the database from lumen plants this little list of matching plants never change during the day jury very stable that that can have very strong cacheability
properties we found that giving front end client more data meant exponentially were less work through our servers end of which was an
early lesson for me been thinking about this problem thinking about this issue of of doubt of very often project and just racing to try to do and obviously there's correspond what the user wants to see and visit and this was the 1st experience in stepping back and thinking about the relationship that the the pipeline that and building as a data pipeline that think of my red that as kind of a Danish often trying to get the chunks of data that makes the most sense and for at the client fast enough that it can put in front of the user the results that the user expects yeah if I'm correct that very often are the solution to building a really responsive really efficient really easily cacheable pop up is thought to put that together in interesting ways I think that is web developers we should have a line on what is happening in this variance in the data space uh because after they have been on a fairly obscure language in many parts of the world for many decades a Python and how that is what I understood that a century ahead is certainly the big breakaway hit on 1 of nanotube breakaway hits in the areas of data processing where larger scientific disciplines of the scientist a very conservative they still use that when a is a very interesting but not comparing Python to Ruby where you can see very often Python is being presented by the younger grad students as a competitor for 20 and the professors are called that and so they shake their heads and they keep writing poetry and talking about how they used to have a machine room full of machines and serve as a bad smaller and more agile sciences for example astronomy because it's a fairly small close-knit community compared to science but knowledge uh has noted within the last 2 or 3 years move over to Python recently I silently Jacob who's doing a lot with the IPython notebook and with science as scientific reproducibility in statistics and assigned to the talk where he showed this slide images of telescopes that are under construction and pointed out that every single 1 of them had an API that's in Python where if you get time meant that research telescope is built on the telescope itself what you given is a Python program that they can go ahead and test before you peer precious few hours on telescope comes up out there is a program that would point the telescope and take the pictures you need and in return for your research grant you use and them on astronomers by the way people who care about the data there right now building an array of telescopes in the southern hemisphere that when complete which would take a few years be generating Exabyte images every day they're having to wonder about things like how do you get enough to power power up to the observatory to reduce the data to the point where it can go back down the mountain every day and because they want to basically take a picture of the entire sky high resolution every 3 days essentially do a digital digital slow us known Sky Survey every 2 or 3 days and that's too much data for them to look at what they're going to do is just give us a pint where we can do whatever research we're not on all of that that's in motion that happens every nite in the sky it the other researchers that have time to build the algorithms that look for near-Earth asteroids work stars momentarily doing is 1 of the planets plant uh crosses in front of the disk or whatever the astronomers are going to be dealing already dealing with big data it's scientists that a pioneer in a lot of work in scientific and consultancies of figuring out how high has created 1 of the simplest possible languages you can control a scientist a grad student is not a programmer doesn't want to be a programmer there busy with astronomy and get them up and running and serving interesting problems I I think that uh behooves us as I've been doing some work recently to just on the side of that about them I erase that in that millions of numbers with a single line of Python code and no apparent loop that hinders data frames essentially having a spreadsheet inside of an object in python with columns and rows of data that you can query and update group by and some with a single method calls but the you're able to do with these vector rooted libraries queries like this it might look a little crazy to read from the inside out person is a data-frame database table but uh in ran on secondary storage A . bg is 1 column of the table and list of phone numbers were imagining in this case get a similar number in Python and saying greater than or equal to 21 year olds which you values true or false so if you get a million ages a conical clan of million ages and compare to 21 you get a non intrusive thoughts In a total of vector of variance and the rule is that you can get either a column right down to the age where table like a person in and does and index it by a list of trees and fences that's is Talis the table is an 6 out the rows for which true is that really in value and so on a single line of code you can get your spreadsheet you die frame then begin to reduce it down to chest the rows you're interested in using a very concise I notation on that has no explicit loops that happened for you behind the scenes and the uh Continuum analytics to spend their life happened sciences and businesses use Python for Big Data has just been writing and
ways which given that syntax and not only use it as it naturally well against number higher those they can also translated into sequence they can also translated into Mandarin DB but can also translated
into Apache Spark or any of the growing list of backends they are say often getting into situations where they need to run an algorithm the familiar with but against a new unfamiliar back in the uh um a customer legacy data is on graphical but the new data they wanna join it with this and that's the actually building not an him I bet the batter mapping engine that lets you ask them to be the same with the and lets you build an object from 1 database well they're letting you right singular lines of code that ask about enormous amounts of data that might be spread across many back or if you're using Apache Spark there might be out shouted across an entire cluster of machines python is
becoming an important tool for large datasets and I think that it would behoove us especially as shown and that the front and becomes more common in many possible to watch how these data to his might fit into
projects in general there's going to be a great web framework to be on top of to be along for this ride with that because it allows such flexibility and randomization providing common patterns of I happen to have an object or page but literally grown Kroger grown well have something more interesting than that a big deal especially on its successes it has no dependencies which for people still want stair-step by hand especially if the organization hours and is that at Enterprise Server This is a big deal for you know houses simply text everyone as gender was competing with thought that you you should mean objects in environment at tributes that gender was the 1st man that was might get the is out of my code let's say that the user is more than the down side of the house and I was very suspicious about at 1st because it felt so right and what guarantees that the such persons slash 21 Mettler T is that a stash person page even exists if it didn't traverse an object on the person number 21 if you from the moment we're the you know as visited a file-system hierarchy it's felt like the intermediate parts of the world had the and somewhere 1st before returning the data chamber really achieved a simplification by saying no animals the text reviews can be simpler procedures it privileges relational databases we should see your gender who much thanks for the fact that generation after generation now pipe on web developer has learned a secret backing and as a the helm many no secret back ends have come and gone over the time gender was being here but people using gender are still around reliable industrial if that's true relational databases by default it is 1 of the default and joining the was the really the 1st web framework I think in pipeline yeah given the fact that finding the simplest possible thing is harder to be the 1st to market with
some real simplicity and I think that solutions can last for quite some time and I'm happy to be along for the ride it but very much 1st I don't know and
Softwaretest
Programmiergerät
Shape <Informatik>
Rahmenproblem
Graph
Güte der Anpassung
Rotationsfläche
Versionsverwaltung
Regulärer Ausdruck
Kolmogorov-Komplexität
Imaginäre Zahl
Kartesische Koordinaten
Trajektorie <Mathematik>
Komplex <Algebra>
Code
Computeranimation
Datensatz
Software
Code
Projektive Ebene
Aggregatzustand
Code
Wasserdampftafel
Mailing-Liste
Optimierung
Raum-Zeit
Code
Computeranimation
Aggregatzustand
Programmiergerät
Punkt
Formale Sprache
Schreiben <Datenverarbeitung>
Komplex <Algebra>
Computeranimation
Eins
Übergang
Cross-site scripting
Fahne <Mathematik>
Code
Mustersprache
Dateiverwaltung
Skript <Programm>
Gerade
Zentrische Streckung
Güte der Anpassung
Algorithmische Programmiersprache
Ereignishorizont
Mustersprache
Verkettung <Informatik>
Verknüpfungsglied
Framework <Informatik>
Geschlecht <Mathematik>
Server
Objektrelationale Abbildung
Zeichenkette
Mathematisierung
Hierarchische Struktur
Framework <Informatik>
Code
Benutzerbeteiligung
Datensatz
Webforum
Flächentheorie
Software
Endogene Variable
Canadian Mathematical Society
Pi <Zahl>
Luenberger-Beobachter
Bildgebendes Verfahren
Attributierte Grammatik
Schreib-Lese-Kopf
Einfach zusammenhängender Raum
Ruby on Rails
Physikalisches System
Regulärer Ausdruck
Objekt <Kategorie>
Softwareschwachstelle
Mereologie
Gamecontroller
Codierung
Web Site
Prozess <Physik>
Momentenproblem
Extrempunkt
Physikalismus
Applet
Gruppenoperation
Extrempunkt
Trajektorie <Mathematik>
Web-Seite
Punktspektrum
Term
Framework <Informatik>
Interrupt <Informatik>
Computeranimation
Übergang
Homepage
Task
Hydrostatik
Festplattenlaufwerk
Client
Benutzerbeteiligung
Suchmaschine
Gruppe <Mathematik>
Endogene Variable
Skript <Programm>
Abstand
Speicher <Informatik>
Touchscreen
Datenhaltung
Gebäude <Mathematik>
Physikalisches System
Bitrate
Elektronische Publikation
Dichte <Physik>
Generator <Informatik>
Framework <Informatik>
Geschlecht <Mathematik>
Offene Menge
Automatische Indexierung
Basisvektor
Mereologie
Debugging
Client
Projektive Ebene
Wort <Informatik>
Normalvektor
Facebook
Mikroprozessor
Prozess <Physik>
Extrempunkt
Güte der Anpassung
Ähnlichkeitsgeometrie
Automatische Differentiation
Computerunterstütztes Verfahren
Elektronische Publikation
Bitrate
Computeranimation
Keller <Informatik>
Homepage
Festplattenlaufwerk
Knotenmenge
Client
Datensatz
Coprozessor
Speicher <Informatik>
Ordnung <Mathematik>
Videospiel
Datenmissbrauch
Prozess <Physik>
Punkt
Momentenproblem
Datenhaltung
Zwei
Versionsverwaltung
Zahlenbereich
Kartesische Koordinaten
Extrempunkt
Rechnen
Inzidenzalgebra
Computeranimation
Homepage
Videokonferenz
Festplattenlaufwerk
Client
Flächeninhalt
Hypermedia
Projektive Ebene
Programmbibliothek
Auswahlaxiom
Lineares Funktional
Sichtenkonzept
Extrempunkt
Gemeinsamer Speicher
Datenhaltung
Versionsverwaltung
Ausnahmebehandlung
Kartesische Koordinaten
Extrempunkt
Sichtenkonzept
Punktspektrum
Framework <Informatik>
Computeranimation
Projektive Ebene
Information
Beobachtungsstudie
Prozess <Physik>
Schlüsselverwaltung
Systemidentifikation
t-Test
Gruppenkeim
Dichotomie
Verzweigendes Programm
EDV-Beratung
Zahlenbereich
Computeranimation
Geschlecht <Mathematik>
Heegaard-Zerlegung
Projektive Ebene
Energieerhaltung
Softwareentwickler
Schlüsselverwaltung
ART-Netz
Expertensystem
Web Site
Momentenproblem
Kategorie <Mathematik>
t-Test
Reihe
Gruppenkeim
Dichotomie
Nummerung
Abenteuerspiel
Computeranimation
Übergang
Homepage
Netzwerktopologie
Gruppenkeim
Flächentheorie
Inverser Limes
Schlüsselverwaltung
ART-Netz
Familie <Mathematik>
Resultante
Soundverarbeitung
Softwaretest
Güte der Anpassung
Digitalfilter
Physikalisches System
Knotenmenge
Computeranimation
Konfiguration <Informatik>
Spannweite <Stochastik>
Multiplikation
Menge
ART-Netz
Projektive Ebene
ART-Netz
Auswahlaxiom
Touchscreen
Resultante
CDN-Netzwerk
Gewicht <Mathematik>
Prozess <Physik>
Gruppenoperation
Applet
Gruppenkeim
t-Test
Iteration
E-Mail
Computeranimation
Homepage
Eins
Mailing-Liste
Benutzerbeteiligung
Iteration
Skript <Programm>
Inhalt <Mathematik>
E-Mail
Bijektion
ART-Netz
Analogieschluss
Touchscreen
Leistungsbewertung
Caching
Systemaufruf
Abfrage
Mailing-Liste
Gruppenoperation
Last
Rechter Winkel
ART-Netz
Server
Dateiformat
Tabelle <Informatik>
Aggregatzustand
Arithmetisches Mittel
Subtraktion
Filter <Stochastik>
Adressraum
Abfrage
Kartesische Koordinaten
Faktor <Algebra>
Computeranimation
Resultante
Client
Filter <Stochastik>
Applet
Anwendungsspezifischer Prozessor
Versionsverwaltung
Skript <Programm>
Mailing-Liste
Digitalfilter
Ordnung <Mathematik>
Computeranimation
Richtung
Resultante
CDN-Netzwerk
Subtraktion
Kontrollstruktur
Hausdorff-Dimension
Wärmeübergang
Mathematische Logik
Computeranimation
Informationsmodellierung
Minimum
Abschattung
Thread
Inhalt <Mathematik>
Integraloperator
E-Mail
Caching
App <Programm>
Sichtenkonzept
Datenhaltung
Abfrage
Auswahlverfahren
CDN-Netzwerk
Rechter Winkel
Mereologie
Debugging
Server
Web-Designer
Standardabweichung
Caching
CDN-Netzwerk
Web Site
Kontrollstruktur
Spider <Programm>
Versionsverwaltung
Web-Seite
Binder <Informatik>
Raum-Zeit
Computeranimation
Homepage
Client
Benutzerbeteiligung
Generator <Informatik>
Mittelwert
Server
Bandmatrix
Bandmatrix
Normalvektor
Aggregatzustand
Lesen <Datenverarbeitung>
Caching
Filter <Stochastik>
Gruppenkeim
t-Test
Abfrage
Digitalfilter
Computerunterstützte Übersetzung
Computeranimation
URL
Schreib-Lese-Kopf
Homepage
Subtraktion
Bit
t-Test
Familie <Mathematik>
Zahlenbereich
Fortsetzung <Mathematik>
Binärcode
Raum-Zeit
Computeranimation
Übergang
Minimum
Softwareentwickler
Drei
Bildgebendes Verfahren
ART-Netz
Shape <Informatik>
Filter <Stochastik>
Datenhaltung
Gebäude <Mathematik>
Relativitätstheorie
Netzwerktopologie
Verbandstheorie
Phasenumwandlung
Verbandstheorie
Information
Binäre Relation
Aggregatzustand
Tabelle <Informatik>
Filter <Stochastik>
Befehl <Informatik>
Datenhaltung
Pivot-Operation
Klasse <Mathematik>
Pivot-Operation
Mailing-Liste
Paarvergleich
Computerunterstütztes Verfahren
Computeranimation
Netzwerktopologie
Mailing-Liste
Client
Gleichheitszeichen
Softwareentwickler
ART-Netz
Tabelle <Informatik>
Shape <Informatik>
Vervollständigung <Mathematik>
Datenhaltung
Applet
Glättung
Wurm <Informatik>
Mailing-Liste
Quick-Sort
Abenteuerspiel
Raum-Zeit
Computeranimation
Konfiguration <Informatik>
Netzwerktopologie
Client
Client
Server
Skript <Programm>
ART-Netz
Bildgebendes Verfahren
Gerade
Aggregatzustand
Prozess <Physik>
Wasserdampftafel
t-Test
Zahlenbereich
Computerunterstütztes Verfahren
E-Mail
Code
Computeranimation
Mailing-Liste
Prozess <Informatik>
Skript <Programm>
Gerade
ART-Netz
Nichtlinearer Operator
Shape <Informatik>
Einfache Genauigkeit
Mailing-Liste
Ähnlichkeitsgeometrie
Bitrate
Ganze Zahl
Debugging
Projektive Ebene
URL
Kantenfärbung
Information
Ordnung <Mathematik>
Aggregatzustand
Resultante
Programmiergerät
Punkt
Formale Sprache
t-Test
Sondierung
Raum-Zeit
Computeranimation
Gradient
Netzwerktopologie
Client
Algorithmus
Datenverarbeitung
Gerade
Bildauflösung
Softwaretest
Konstruktor <Informatik>
Statistik
Kategorie <Mathematik>
Datenhaltung
Gebäude <Mathematik>
Abfrage
Ähnlichkeitsgeometrie
Ausnahmebehandlung
Rechenschieber
Tabellenkalkulation
Automatische Indexierung
Client
Server
Projektive Ebene
Tabelle <Informatik>
Server
Total <Mathematik>
Rahmenproblem
EDV-Beratung
Zahlenbereich
Analytische Menge
Haftreibung
Wissenstechnik
Code
Demoszene <Programmierung>
Virtuelle Maschine
Loop
Datensatz
Kugel
Mini-Disc
Notebook-Computer
Programmbibliothek
Optimierung
Speicher <Informatik>
Ganze Funktion
Bildgebendes Verfahren
Varianz
URL
Schreib-Lese-Kopf
Leistung <Physik>
Videospiel
Kontinuumshypothese
Einfache Genauigkeit
Schlussregel
Mailing-Liste
Vektorraum
Objekt <Kategorie>
Array <Informatik>
Flächeninhalt
Debugging
Mereologie
Web-Designer
Boolesche Algebra
Folge <Mathematik>
Datenhaltung
Zahlenbereich
Mailing-Liste
Code
Computeranimation
Objekt <Kategorie>
Mapping <Computergraphik>
Singularität <Mathematik>
Virtuelle Maschine
Algorithmus
Rechter Winkel
Front-End <Software>
Ganze Funktion
Gerade
Relationale Datenbank
Momentenproblem
Selbst organisierendes System
Zahlenbereich
Hierarchische Struktur
Framework <Informatik>
Code
Computeranimation
Datenhaltung
Homepage
Benutzerbeteiligung
Polygonzug
Code
Mustersprache
Randomisierung
Dateiverwaltung
Default
URL
Metropolitan area network
Relationale Datenbank
Menge
Algorithmische Programmiersprache
Arithmetisches Mittel
Objekt <Kategorie>
Generator <Informatik>
Geschlecht <Mathematik>
Mereologie
Server
Web-Designer
Projektive Ebene
Programmierumgebung
Vorlesung/Konferenz
Ereignishorizont
Computeranimation

Metadaten

Formale Metadaten

Titel Django: A Data Shovel With a Future
Serientitel DjangoCon US 2014
Teil 33
Anzahl der Teile 44
Autor Rhodes, Brandon
Mitwirkende Confreaks, LLC
Lizenz CC-Namensnennung - Weitergabe unter gleichen Bedingungen 4.0 International:
Sie dürfen das Werk bzw. den Inhalt zu jedem legalen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen und das Werk bzw. diesen Inhalt auch in veränderter Form nur unter den Bedingungen dieser Lizenz weitergeben.
DOI 10.5446/32825
Herausgeber DjangoCon US
Erscheinungsjahr 2014
Sprache Englisch

Inhaltliche Metadaten

Fachgebiet Informatik

Ähnliche Filme

Loading...