Up next in 10
The Code Quality Advantage How Empirical Data Shatters the Speed vs Quality Myth
0 views
Aug 6, 2025
Code quality is an abstract concept that fails to get traction at the business level. Consequently, software companies keep trading code quality for new features. The resulting technical debt is estimated to waste up to 42% of developers' time, causing stress and uncertainty in the process. Yet it's hard to build a business case for code quality: how do we quantify and communicate the benefits to our non-technical stakeholders? Or even inside the own team? In this talk, Adam tackles this challenge head-on by leveraging innovative code quality metrics and deep analyses of how engineering interacts with code. By linking these metrics to key business drivers such as time-to-market, customer satisfaction, and roadmap risks, Adam demonstrates how investing in code quality translates directly into a competitive advantage. The recommendations and insights are backed by cutting-edge research, delivering a perspective on software development that will change how you view code. Promise. š Code Quality Conference 2024 https://codequalityconf.com/ šŗ CSharp TV - Dev Streaming Destination http://csharp.tv š C# Corner - Community of Software and Data Developers https://www.c-sharpcorner.com/ #CSharpTV #csharpcorner #CodeQualityConf
View Video Transcript
0:03
and welcome everyone to this session on
0:05
and welcome everyone to this session on
0:05
and welcome everyone to this session on the code quality Advantage so the reason
0:08
the code quality Advantage so the reason
0:08
the code quality Advantage so the reason I want to do this session is because
0:10
I want to do this session is because
0:10
I want to do this session is because I've been spending 30 years in the
0:13
I've been spending 30 years in the
0:13
I've been spending 30 years in the software industry been writing a lot of
0:15
software industry been writing a lot of
0:15
software industry been writing a lot of code work with lots of different
0:16
code work with lots of different
0:16
code work with lots of different companies and
0:17
companies and
0:17
companies and stakeholders and one common theme that I
0:20
stakeholders and one common theme that I
0:20
stakeholders and one common theme that I noticed is that people during these 30
0:22
noticed is that people during these 30
0:22
noticed is that people during these 30 years have been telling me no no we
0:26
years have been telling me no no we
0:26
years have been telling me no no we don't have time to refactor no we don't
0:28
don't have time to refactor no we don't
0:28
don't have time to refactor no we don't have time to automate tests no we don't
0:30
have time to automate tests no we don't
0:30
have time to automate tests no we don't have time to rethink our architecture so
0:34
have time to rethink our architecture so
0:34
have time to rethink our architecture so we have this misconception in the
0:36
we have this misconception in the
0:36
we have this misconception in the industry that there is some kind of
0:37
industry that there is some kind of
0:37
industry that there is some kind of tradeoff between speed and quality right
0:41
tradeoff between speed and quality right
0:41
tradeoff between speed and quality right that we need to sacrifice one to get the
0:43
that we need to sacrifice one to get the
0:43
that we need to sacrifice one to get the other and what I want to show you today
0:45
other and what I want to show you today
0:45
other and what I want to show you today is that this is a myth there is no such
0:47
is that this is a myth there is no such
0:47
is that this is a myth there is no such tradeoff in fact we need both to go
0:50
tradeoff in fact we need both to go
0:50
tradeoff in fact we need both to go really really quick and we're going to
0:52
really really quick and we're going to
0:53
really really quick and we're going to cover a lot of ground over the next 40
0:55
cover a lot of ground over the next 40
0:55
cover a lot of ground over the next 40 minutes and I like to approach this from
0:58
minutes and I like to approach this from
0:58
minutes and I like to approach this from the perspective of technical that
1:01
the perspective of technical that
1:01
the perspective of technical that so technical left is something that you
1:03
so technical left is something that you
1:04
so technical left is something that you probably are all familiar with you have
1:05
probably are all familiar with you have
1:05
probably are all familiar with you have probably heard the term before what I
1:08
probably heard the term before what I
1:08
probably heard the term before what I think is so interesting is that
1:09
think is so interesting is that
1:10
think is so interesting is that technical depth has multiple root causes
1:12
technical depth has multiple root causes
1:12
technical depth has multiple root causes we take it on for multiple reasons and
1:15
we take it on for multiple reasons and
1:15
we take it on for multiple reasons and the typical case is that we try to
1:19
the typical case is that we try to
1:19
the typical case is that we try to sacrifice quality in order to move a
1:21
sacrifice quality in order to move a
1:21
sacrifice quality in order to move a little bit faster with a feature right
1:24
little bit faster with a feature right
1:24
little bit faster with a feature right so we cut some Corners we get a feature
1:26
so we cut some Corners we get a feature
1:26
so we cut some Corners we get a feature out in production and everyone is happy
1:28
out in production and everyone is happy
1:28
out in production and everyone is happy in the short term
1:30
in the short term
1:30
in the short term however there are other reasons too that
1:32
however there are other reasons too that
1:32
however there are other reasons too that we take on technical that and one of the
1:34
we take on technical that and one of the
1:35
we take on technical that and one of the reasons might be that we actually do the
1:37
reasons might be that we actually do the
1:37
reasons might be that we actually do the right thing we design our software
1:39
right thing we design our software
1:39
right thing we design our software properly but then our understanding of
1:42
properly but then our understanding of
1:42
properly but then our understanding of the problem we're trying to solve
1:43
the problem we're trying to solve
1:43
the problem we're trying to solve changes perhaps due to user feedback of
1:46
changes perhaps due to user feedback of
1:46
changes perhaps due to user feedback of our own experience and this is what we
1:48
our own experience and this is what we
1:48
our own experience and this is what we typically call requirements change
1:51
typically call requirements change
1:51
typically call requirements change right and now our design is no longer
1:53
right and now our design is no longer
1:53
right and now our design is no longer good fit or maybe and of course highly
1:57
good fit or maybe and of course highly
1:57
good fit or maybe and of course highly hypothetical in your case maybe our
1:59
hypothetical in your case maybe our
1:59
hypothetical in your case maybe our design just wasn't a good fit to start
2:01
design just wasn't a good fit to start
2:01
design just wasn't a good fit to start with right no matter the root cause the
2:05
with right no matter the root cause the
2:05
with right no matter the root cause the outcome is the same we end up with code
2:07
outcome is the same we end up with code
2:07
outcome is the same we end up with code that is more expensive to maintain than
2:09
that is more expensive to maintain than
2:09
that is more expensive to maintain than it should be and this is the definition
2:12
it should be and this is the definition
2:12
it should be and this is the definition of technical depth that I'm going to run
2:14
of technical depth that I'm going to run
2:14
of technical depth that I'm going to run with throughout this
2:16
with throughout this
2:16
with throughout this presentation now during the past four to
2:19
presentation now during the past four to
2:19
presentation now during the past four to five years we have learned a lot about
2:21
five years we have learned a lot about
2:21
five years we have learned a lot about the industry cost of bad code and
2:23
the industry cost of bad code and
2:23
the industry cost of bad code and Technical depth in general and there are
2:26
Technical depth in general and there are
2:26
Technical depth in general and there are multiple papers you can look at I'm
2:28
multiple papers you can look at I'm
2:28
multiple papers you can look at I'm going to reference a few of them here
2:30
going to reference a few of them here
2:30
going to reference a few of them here but at the end of the day it doesn't
2:32
but at the end of the day it doesn't
2:32
but at the end of the day it doesn't matter which paper you read because they
2:34
matter which paper you read because they
2:34
matter which paper you read because they have one thing in common they all paint
2:36
have one thing in common they all paint
2:36
have one thing in common they all paint a pretty depressing view of our
2:38
a pretty depressing view of our
2:38
a pretty depressing view of our industry did you know that as developers
2:41
industry did you know that as developers
2:41
industry did you know that as developers we waste somewhere between 23 to
2:45
we waste somewhere between 23 to
2:45
we waste somewhere between 23 to 42% of our work week dealing with the
2:47
42% of our work week dealing with the
2:47
42% of our work week dealing with the consequences of technical depth and bad
2:49
consequences of technical depth and bad
2:49
consequences of technical depth and bad code in
2:51
code in
2:51
code in general 42% waste what does that even
2:55
general 42% waste what does that even
2:55
general 42% waste what does that even mean let's do a small F experiment
2:58
mean let's do a small F experiment
2:58
mean let's do a small F experiment together so let's pretend that you have
3:01
together so let's pretend that you have
3:01
together so let's pretend that you have a company with 100 Engineers that means
3:04
a company with 100 Engineers that means
3:04
a company with 100 Engineers that means you need to pay the salaries for 100
3:06
you need to pay the salaries for 100
3:06
you need to pay the salaries for 100 people with a 42% waste it means you pay
3:10
people with a 42% waste it means you pay
3:10
people with a 42% waste it means you pay the salaries for 100 people but you get
3:12
the salaries for 100 people but you get
3:12
the salaries for 100 people but you get the equivalent output of just 58 people
3:15
the equivalent output of just 58 people
3:15
the equivalent output of just 58 people does that sound like a good deal to you
3:18
does that sound like a good deal to you
3:18
does that sound like a good deal to you no of course not it's
3:20
no of course not it's
3:20
no of course not it's horrible and of course there are other
3:23
horrible and of course there are other
3:23
horrible and of course there are other factors too because I like to claim that
3:25
factors too because I like to claim that
3:26
factors too because I like to claim that if you could actually build something
3:27
if you could actually build something
3:27
if you could actually build something with 58 people as opposed to 100
3:30
with 58 people as opposed to 100
3:30
with 58 people as opposed to 100 then doing it with the smaller team is
3:32
then doing it with the smaller team is
3:32
then doing it with the smaller team is always always going to be more efficient
3:34
always always going to be more efficient
3:35
always always going to be more efficient because with a larger team you have
3:36
because with a larger team you have
3:36
because with a larger team you have additional coordination costs
3:38
additional coordination costs
3:38
additional coordination costs communication costs and management
3:40
communication costs and management
3:40
communication costs and management overhead so I actually think that the
3:42
overhead so I actually think that the
3:43
overhead so I actually think that the real waste from technical debt is
3:44
real waste from technical debt is
3:44
real waste from technical debt is significantly larger I think these
3:47
significantly larger I think these
3:47
significantly larger I think these numbers are overly
3:49
numbers are overly
3:49
numbers are overly optimistic of course technical depth and
3:52
optimistic of course technical depth and
3:52
optimistic of course technical depth and bad code is not only about financial
3:54
bad code is not only about financial
3:55
bad code is not only about financial impact it's also about vulnerabilities
3:58
impact it's also about vulnerabilities
3:58
impact it's also about vulnerabilities so another line of research has
3:59
so another line of research has
3:59
so another line of research has established that there is a strong
4:02
established that there is a strong
4:02
established that there is a strong correlation between software
4:04
correlation between software
4:04
correlation between software vulnerabilities and Technical depth in
4:07
vulnerabilities and Technical depth in
4:07
vulnerabilities and Technical depth in the shape of various codes Mells that
4:08
the shape of various codes Mells that
4:09
the shape of various codes Mells that we're going to see throughout this
4:11
we're going to see throughout this
4:11
we're going to see throughout this presentation so now where does this
4:13
presentation so now where does this
4:13
presentation so now where does this leave us
4:15
leave us
4:15
leave us well if we have something technical
4:17
well if we have something technical
4:17
well if we have something technical depth bad code that uh has a massive
4:20
depth bad code that uh has a massive
4:20
depth bad code that uh has a massive Financial impact waste for the 2% of our
4:22
Financial impact waste for the 2% of our
4:22
Financial impact waste for the 2% of our time and causes secured
4:24
time and causes secured
4:24
time and causes secured vulnerabilities we would expect
4:26
vulnerabilities we would expect
4:26
vulnerabilities we would expect technical dep to be at the very top of
4:28
technical dep to be at the very top of
4:28
technical dep to be at the very top of any decision maker
4:30
any decision maker
4:30
any decision maker agenda is that what's
4:34
agenda is that what's
4:34
agenda is that what's happening well probably not because what
4:37
happening well probably not because what
4:37
happening well probably not because what again what we see in the research field
4:39
again what we see in the research field
4:39
again what we see in the research field is that we do a horrible job as an
4:41
is that we do a horrible job as an
4:42
is that we do a horrible job as an industry in fact what researchers find
4:45
industry in fact what researchers find
4:45
industry in fact what researchers find is that we developers are frequently
4:47
is that we developers are frequently
4:47
is that we developers are frequently forced to introduce even more technical
4:49
forced to introduce even more technical
4:49
forced to introduce even more technical depth because companies just keep
4:51
depth because companies just keep
4:51
depth because companies just keep trading code quality for short-term
4:53
trading code quality for short-term
4:53
trading code quality for short-term gains like the next big
4:56
gains like the next big
4:56
gains like the next big feature now i' I've been thinking about
4:58
feature now i' I've been thinking about
4:58
feature now i' I've been thinking about this for a long time for at least the
5:00
this for a long time for at least the
5:00
this for a long time for at least the past 10 years why do we do this tradeoff
5:03
past 10 years why do we do this tradeoff
5:03
past 10 years why do we do this tradeoff why do we sacrifice code quality over
5:05
why do we sacrifice code quality over
5:05
why do we sacrifice code quality over and over again to ship
5:07
and over again to ship
5:07
and over again to ship features and I think I have an IDE on
5:10
features and I think I have an IDE on
5:11
features and I think I have an IDE on one of the root
5:12
one of the root
5:12
one of the root causes so let me show an example to
5:15
causes so let me show an example to
5:15
causes so let me show an example to you I didn't bring any c code in this
5:18
you I didn't bring any c code in this
5:18
you I didn't bring any c code in this slide I decided to go a little bit more
5:22
slide I decided to go a little bit more
5:22
slide I decided to go a little bit more old school and present a piece of
5:24
old school and present a piece of
5:24
old school and present a piece of C++ but it doesn't really matter right
5:27
C++ but it doesn't really matter right
5:27
C++ but it doesn't really matter right because I think as C Developers familiar
5:30
because I think as C Developers familiar
5:30
because I think as C Developers familiar with that
5:30
with that
5:31
with that structure and I'd like to ask you a
5:33
structure and I'd like to ask you a
5:33
structure and I'd like to ask you a question looking at that code can anyone
5:37
question looking at that code can anyone
5:37
question looking at that code can anyone tell me if this code is correct can
5:40
tell me if this code is correct can
5:40
tell me if this code is correct can anyone tell me what this code does I
5:43
anyone tell me what this code does I
5:43
anyone tell me what this code does I have to admit that after all these years
5:45
have to admit that after all these years
5:45
have to admit that after all these years coding and I even spent 10 years doing
5:47
coding and I even spent 10 years doing
5:47
coding and I even spent 10 years doing C++ I have absolutely no idea what this
5:50
C++ I have absolutely no idea what this
5:50
C++ I have absolutely no idea what this piece of code does and I cannot tell if
5:51
piece of code does and I cannot tell if
5:51
piece of code does and I cannot tell if it's correct or not without spending a
5:54
it's correct or not without spending a
5:54
it's correct or not without spending a lot of time on
5:55
lot of time on
5:55
lot of time on it now what I want to point out is that
5:58
it now what I want to point out is that
5:58
it now what I want to point out is that if we as develop velers struggle to
6:00
if we as develop velers struggle to
6:00
if we as develop velers struggle to understand code we have to keep in mind
6:02
understand code we have to keep in mind
6:02
understand code we have to keep in mind that what you see on screen is just 19
6:04
that what you see on screen is just 19
6:04
that what you see on screen is just 19 lines of code a modern system has
6:08
lines of code a modern system has
6:08
lines of code a modern system has hundreds of thousands million lines of
6:10
hundreds of thousands million lines of
6:10
hundreds of thousands million lines of code that looks like just like that and
6:13
code that looks like just like that and
6:13
code that looks like just like that and if we as developers struggle with
6:15
if we as developers struggle with
6:15
if we as developers struggle with understanding small pieces of code
6:17
understanding small pieces of code
6:17
understanding small pieces of code imagine a non-technical stakeholder like
6:19
imagine a non-technical stakeholder like
6:19
imagine a non-technical stakeholder like a product person a technical manager or
6:22
a product person a technical manager or
6:22
a product person a technical manager or vpf engineering a business
6:25
vpf engineering a business
6:25
vpf engineering a business manager how should they understand if
6:27
manager how should they understand if
6:27
manager how should they understand if our code is good or bad if there
6:29
our code is good or bad if there
6:29
our code is good or bad if there technical depth or where ways this
6:32
technical depth or where ways this
6:32
technical depth or where ways this software is lared all blackbox to many
6:34
software is lared all blackbox to many
6:34
software is lared all blackbox to many many people so what I like us to do
6:37
many people so what I like us to do
6:37
many people so what I like us to do today is to see can we shine a light
6:39
today is to see can we shine a light
6:39
today is to see can we shine a light into this black box can we shine a light
6:41
into this black box can we shine a light
6:41
into this black box can we shine a light on technical dep and make it relevant to
6:43
on technical dep and make it relevant to
6:43
on technical dep and make it relevant to the
6:45
business in order to do this I would
6:48
business in order to do this I would
6:48
business in order to do this I would like us to start to visualize technical
6:50
like us to start to visualize technical
6:50
like us to start to visualize technical LA and code quality and to visualize
6:53
LA and code quality and to visualize
6:53
LA and code quality and to visualize something we need to know what to
6:55
something we need to know what to
6:55
something we need to know what to visualize so we need to know how to
6:57
visualize so we need to know how to
6:57
visualize so we need to know how to measure code quality and this has been a
7:00
measure code quality and this has been a
7:00
measure code quality and this has been a notoriously hard problem there have been
7:03
notoriously hard problem there have been
7:03
notoriously hard problem there have been multiple attempts throughout the decades
7:05
multiple attempts throughout the decades
7:05
multiple attempts throughout the decades and you might have heard about
7:06
and you might have heard about
7:06
and you might have heard about approaches like cycl medic complexity
7:08
approaches like cycl medic complexity
7:08
approaches like cycl medic complexity has its volume metrics and so on and
7:11
has its volume metrics and so on and
7:11
has its volume metrics and so on and what all of these metrics have in common
7:12
what all of these metrics have in common
7:12
what all of these metrics have in common is that there are pretty poor predictors
7:14
is that there are pretty poor predictors
7:14
is that there are pretty poor predictors of complexity in fact there's no real
7:17
of complexity in fact there's no real
7:17
of complexity in fact there's no real correlation between what these metrics
7:19
correlation between what these metrics
7:19
correlation between what these metrics find and something that's meaningful to
7:21
find and something that's meaningful to
7:21
find and something that's meaningful to the
7:22
the
7:22
the business so one thing that uh me and my
7:26
business so one thing that uh me and my
7:26
business so one thing that uh me and my team have been working on for the past
7:27
team have been working on for the past
7:27
team have been working on for the past six to seven years is to see can we
7:29
six to seven years is to see can we
7:29
six to seven years is to see can we develop a better proxy for code
7:32
develop a better proxy for code
7:32
develop a better proxy for code quality and we decided to call the
7:34
quality and we decided to call the
7:34
quality and we decided to call the concept code health so code health is
7:36
concept code health so code health is
7:36
concept code health so code health is both a concept and a specific software
7:38
both a concept and a specific software
7:38
both a concept and a specific software metric that you can use and it's a
7:41
metric that you can use and it's a
7:41
metric that you can use and it's a metric that's based on 25 different
7:43
metric that's based on 25 different
7:43
metric that's based on 25 different factors why 25
7:46
factors why 25
7:46
factors why 25 factors well the reason is simple it's
7:49
factors well the reason is simple it's
7:49
factors well the reason is simple it's very hard to agree on what makes good
7:52
very hard to agree on what makes good
7:52
very hard to agree on what makes good code good right it's very hard to Define
7:55
code good right it's very hard to Define
7:55
code good right it's very hard to Define quality but it's very easy to Define
7:58
quality but it's very easy to Define
7:58
quality but it's very easy to Define what bad code is so what we did is that
8:00
what bad code is so what we did is that
8:00
what bad code is so what we did is that we simply researched and came up with a
8:02
we simply researched and came up with a
8:02
we simply researched and came up with a list of 25 factors that we know are
8:05
list of 25 factors that we know are
8:05
list of 25 factors that we know are indicators of bad code so the way this
8:08
indicators of bad code so the way this
8:08
indicators of bad code so the way this works is that you have a piece of code
8:10
works is that you have a piece of code
8:10
works is that you have a piece of code might be C might be yava or something
8:13
might be C might be yava or something
8:13
might be C might be yava or something else and you parse that code and then
8:16
else and you parse that code and then
8:16
else and you parse that code and then you Analyze That code for these
8:18
you Analyze That code for these
8:18
you Analyze That code for these different codes smells and I'm not going
8:20
different codes smells and I'm not going
8:20
different codes smells and I'm not going to show you the full list of the 25
8:22
to show you the full list of the 25
8:22
to show you the full list of the 25 factors you free to check it out uh via
8:24
factors you free to check it out uh via
8:25
factors you free to check it out uh via the links but I'm going to show you the
8:26
the links but I'm going to show you the
8:26
the links but I'm going to show you the most important ones so at the top level
8:30
most important ones so at the top level
8:31
most important ones so at the top level we look for um module level smells so
8:34
we look for um module level smells so
8:34
we look for um module level smells so these are like class design issues and
8:36
these are like class design issues and
8:36
these are like class design issues and to give you a typical example of a class
8:38
to give you a typical example of a class
8:38
to give you a typical example of a class design issue or code smell at that level
8:41
design issue or code smell at that level
8:41
design issue or code smell at that level that's low
8:43
that's low
8:43
that's low cohesion so you might be familiar with
8:45
cohesion so you might be familiar with
8:45
cohesion so you might be familiar with low cohesion it basically means that you
8:47
low cohesion it basically means that you
8:47
low cohesion it basically means that you have stuffed too many business
8:49
have stuffed too many business
8:49
have stuffed too many business responsibilities into the same class so
8:52
responsibilities into the same class so
8:52
responsibilities into the same class so now this class has many many reasons to
8:54
now this class has many many reasons to
8:54
now this class has many many reasons to change and this is problematic because
8:56
change and this is problematic because
8:56
change and this is problematic because as a developer trying to understand that
8:58
as a developer trying to understand that
8:58
as a developer trying to understand that code I now need need to understand all
9:00
code I now need need to understand all
9:00
code I now need need to understand all these different business uh rules in
9:02
these different business uh rules in
9:02
these different business uh rules in order to make sense of the code it also
9:05
order to make sense of the code it also
9:05
order to make sense of the code it also opens up the risk for things like
9:07
opens up the risk for things like
9:07
opens up the risk for things like unexpected feature interactions which
9:09
unexpected feature interactions which
9:09
unexpected feature interactions which are some of the worst bugs you can have
9:11
are some of the worst bugs you can have
9:11
are some of the worst bugs you can have you know you tweak some code over here
9:13
you know you tweak some code over here
9:13
you know you tweak some code over here and a completely unrelated feature
9:15
and a completely unrelated feature
9:15
and a completely unrelated feature breaks that's usually caused by low
9:18
breaks that's usually caused by low
9:18
breaks that's usually caused by low cohesion so low cohesion is a very
9:20
cohesion so low cohesion is a very
9:20
cohesion so low cohesion is a very serious smell it obviously doesn't mean
9:23
serious smell it obviously doesn't mean
9:23
serious smell it obviously doesn't mean that we cannot make it even worse
9:25
that we cannot make it even worse
9:25
that we cannot make it even worse because what we can do is we can take a
9:26
because what we can do is we can take a
9:26
because what we can do is we can take a class with low cohesion and let it grow
9:28
class with low cohesion and let it grow
9:28
class with low cohesion and let it grow really really l and let it include at
9:30
really really l and let it include at
9:30
really really l and let it include at least a brain method I'm going to talk
9:32
least a brain method I'm going to talk
9:32
least a brain method I'm going to talk about that soon what we have done now is
9:34
about that soon what we have done now is
9:34
about that soon what we have done now is that we created a brain class so that's
9:36
that we created a brain class so that's
9:36
that we created a brain class so that's another modu level
9:38
another modu level
9:38
another modu level smell and brain methods are a function
9:41
smell and brain methods are a function
9:41
smell and brain methods are a function level smell it's something you might
9:43
level smell it's something you might
9:43
level smell it's something you might also known under its Alias gut
9:46
also known under its Alias gut
9:46
also known under its Alias gut functions what brain methods and gut
9:48
functions what brain methods and gut
9:48
functions what brain methods and gut functions are are simply large
9:51
functions are are simply large
9:51
functions are are simply large functions they are always large and they
9:54
functions they are always large and they
9:54
functions they are always large and they tend to contain a lot of logic and the
9:56
tend to contain a lot of logic and the
9:56
tend to contain a lot of logic and the reason they contain a lot of logic is
9:57
reason they contain a lot of logic is
9:57
reason they contain a lot of logic is because again they do two to many things
10:00
because again they do two to many things
10:00
because again they do two to many things they are again low on cohesion so what
10:02
they are again low on cohesion so what
10:03
they are again low on cohesion so what this means to me as a developer is that
10:04
this means to me as a developer is that
10:04
this means to me as a developer is that the moment I want to modify anything
10:07
the moment I want to modify anything
10:07
the moment I want to modify anything inside this particular area of the code
10:08
inside this particular area of the code
10:08
inside this particular area of the code end up in the brain method and each time
10:11
end up in the brain method and each time
10:11
end up in the brain method and each time I work on it it becomes more and more
10:13
I work on it it becomes more and more
10:13
I work on it it becomes more and more complicated so brain methods is a pretty
10:16
complicated so brain methods is a pretty
10:16
complicated so brain methods is a pretty severe code
10:17
severe code
10:17
severe code smell then finally at the implementation
10:20
smell then finally at the implementation
10:20
smell then finally at the implementation level we look for various complexity
10:23
level we look for various complexity
10:23
level we look for various complexity patterns and um to mention a pattern
10:26
patterns and um to mention a pattern
10:26
patterns and um to mention a pattern with really high predictive power I like
10:29
with really high predictive power I like
10:29
with really high predictive power I like to point out deep nested logic so deep
10:32
to point out deep nested logic so deep
10:33
to point out deep nested logic so deep nested logic is something all of you
10:34
nested logic is something all of you
10:34
nested logic is something all of you have seen it's when you have an if
10:36
have seen it's when you have an if
10:36
have seen it's when you have an if statement inside an if statement inside
10:38
statement inside an if statement inside
10:38
statement inside an if statement inside an if statement and maybe a loop for
10:40
an if statement and maybe a loop for
10:40
an if statement and maybe a loop for good
10:41
good
10:41
good measure the reason that code is
10:43
measure the reason that code is
10:43
measure the reason that code is problematic is because there is some
10:45
problematic is because there is some
10:45
problematic is because there is some very good research that shows that
10:47
very good research that shows that
10:47
very good research that shows that roughly 20% of all programming mistakes
10:49
roughly 20% of all programming mistakes
10:49
roughly 20% of all programming mistakes are due to things like deepness and
10:51
are due to things like deepness and
10:51
are due to things like deepness and logic and once we know about this it's
10:54
logic and once we know about this it's
10:54
logic and once we know about this it's it's pretty easy to avoid right so these
10:57
it's pretty easy to avoid right so these
10:57
it's pretty easy to avoid right so these are just some high level examples on um
11:00
are just some high level examples on um
11:00
are just some high level examples on um the 25 factors that you look for in code
11:02
the 25 factors that you look for in code
11:02
the 25 factors that you look for in code health so you basically look for the
11:04
health so you basically look for the
11:04
health so you basically look for the presence of any of these factors in the
11:07
presence of any of these factors in the
11:07
presence of any of these factors in the code and then uh we have algorithms that
11:09
code and then uh we have algorithms that
11:09
code and then uh we have algorithms that can score Aggregate and categorize every
11:12
can score Aggregate and categorize every
11:12
can score Aggregate and categorize every single piece of code as being in one of
11:14
single piece of code as being in one of
11:14
single piece of code as being in one of three
11:16
three
11:16
three categories the code can either be green
11:18
categories the code can either be green
11:18
categories the code can either be green that's healthy code with low risk that's
11:20
that's healthy code with low risk that's
11:20
that's healthy code with low risk that's where you want to be with the majority
11:22
where you want to be with the majority
11:22
where you want to be with the majority of your source
11:24
of your source
11:24
of your source code and then you have the next uh
11:27
code and then you have the next uh
11:27
code and then you have the next uh category yellow code
11:29
category yellow code
11:30
category yellow code that's problematic code where you have
11:31
that's problematic code where you have
11:31
that's problematic code where you have started to take on technical depth and
11:34
started to take on technical depth and
11:34
started to take on technical depth and unless you're really careful you might
11:36
unless you're really careful you might
11:36
unless you're really careful you might end up in the red category which is
11:38
end up in the red category which is
11:38
end up in the red category which is unhealthy code that's the type of code
11:39
unhealthy code that's the type of code
11:39
unhealthy code that's the type of code you never ever want to work with right
11:42
you never ever want to work with right
11:42
you never ever want to work with right that's what typically call Spaghetti
11:43
that's what typically call Spaghetti
11:43
that's what typically call Spaghetti code it's code that's really hard to
11:45
code it's code that's really hard to
11:45
code it's code that's really hard to understand and very fragile as we will
11:47
understand and very fragile as we will
11:47
understand and very fragile as we will soon
11:48
soon
11:48
soon see so now that we know how we can use
11:52
see so now that we know how we can use
11:52
see so now that we know how we can use codal as a proy for code quality we can
11:55
codal as a proy for code quality we can
11:55
codal as a proy for code quality we can start to visualize code quality in large
11:58
start to visualize code quality in large
11:58
start to visualize code quality in large code bases and and I want to show how
12:00
code bases and and I want to show how
12:00
code bases and and I want to show how this can look on a real code
12:02
this can look on a real code
12:02
this can look on a real code base so what you see on screen here is a
12:05
base so what you see on screen here is a
12:05
base so what you see on screen here is a code quality visualization of
12:07
code quality visualization of
12:07
code quality visualization of ephrom and there's really nothing
12:10
ephrom and there's really nothing
12:10
ephrom and there's really nothing special about eom U it's just a good
12:12
special about eom U it's just a good
12:12
special about eom U it's just a good example of a midsized code base and
12:15
example of a midsized code base and
12:15
example of a midsized code base and Ephron for those of you who might not
12:17
Ephron for those of you who might not
12:17
Ephron for those of you who might not know it it's an open source blockchain
12:20
know it it's an open source blockchain
12:20
know it it's an open source blockchain implementation the interesting thing
12:22
implementation the interesting thing
12:22
implementation the interesting thing with this visualization is that it's
12:25
with this visualization is that it's
12:25
with this visualization is that it's based on code written in 50 different
12:28
based on code written in 50 different
12:28
based on code written in 50 different git
12:29
git
12:29
git repositories so it's a multi repo
12:32
repositories so it's a multi repo
12:32
repositories so it's a multi repo project and you see that each one of
12:33
project and you see that each one of
12:34
project and you see that each one of these circles with a name the ones I'm
12:36
these circles with a name the ones I'm
12:36
these circles with a name the ones I'm hovering are right now they correspond
12:38
hovering are right now they correspond
12:38
hovering are right now they correspond to code in a specific git
12:40
to code in a specific git
12:40
to code in a specific git repository if this was a monor repo then
12:44
repository if this was a monor repo then
12:44
repository if this was a monor repo then um the top level circles would
12:45
um the top level circles would
12:45
um the top level circles would correspond to the top level folders in
12:47
correspond to the top level folders in
12:47
correspond to the top level folders in your repositories we're going to see
12:49
your repositories we're going to see
12:49
your repositories we're going to see some examples on that soon as well
12:52
some examples on that soon as well
12:52
some examples on that soon as well anyway uh by using this visualization I
12:54
anyway uh by using this visualization I
12:54
anyway uh by using this visualization I can quickly zoom in on the area I'm
12:55
can quickly zoom in on the area I'm
12:55
can quickly zoom in on the area I'm interested in a typical an area where I
12:57
interested in a typical an area where I
12:57
interested in a typical an area where I do a lot of work or I might an
12:59
do a lot of work or I might an
12:59
do a lot of work or I might an additional technical
13:00
additional technical
13:00
additional technical responsibility so let's zoom in on
13:02
responsibility so let's zoom in on
13:02
responsibility so let's zoom in on Research here so when I zoom in on
13:04
Research here so when I zoom in on
13:04
Research here so when I zoom in on Research what I see is that it's a
13:07
Research what I see is that it's a
13:07
Research what I see is that it's a hierarchical structure so we have the
13:09
hierarchical structure so we have the
13:09
hierarchical structure so we have the various subfolders inside that part of
13:11
various subfolders inside that part of
13:11
various subfolders inside that part of the code and once I get to the lowest
13:13
the code and once I get to the lowest
13:13
the code and once I get to the lowest level of detail I see that each file
13:15
level of detail I see that each file
13:15
level of detail I see that each file with source code is visualized as a
13:18
with source code is visualized as a
13:18
with source code is visualized as a circle and you also see that these
13:20
circle and you also see that these
13:20
circle and you also see that these circles they have different size some
13:22
circles they have different size some
13:22
circles they have different size some are big some are much smaller so the
13:24
are big some are much smaller so the
13:24
are big some are much smaller so the size of a circle just reflects the
13:26
size of a circle just reflects the
13:26
size of a circle just reflects the number of lines of code in that class
13:30
number of lines of code in that class
13:30
number of lines of code in that class the important that's just to know how
13:32
the important that's just to know how
13:32
the important that's just to know how big a potential problem is right the
13:35
big a potential problem is right the
13:35
big a potential problem is right the interesting thing here is the color
13:37
interesting thing here is the color
13:37
interesting thing here is the color which indicates the code ofal so we see
13:39
which indicates the code ofal so we see
13:39
which indicates the code ofal so we see that on the top we have one big red blob
13:41
that on the top we have one big red blob
13:41
that on the top we have one big red blob with red problematic code and below it
13:44
with red problematic code and below it
13:44
with red problematic code and below it we have two healthy pieces of
13:46
we have two healthy pieces of
13:46
we have two healthy pieces of code so this makes it really really easy
13:49
code so this makes it really really easy
13:49
code so this makes it really really easy to pinpoint the unhealthy problematic
13:51
to pinpoint the unhealthy problematic
13:51
to pinpoint the unhealthy problematic parts of your
13:53
parts of your
13:53
parts of your code now I'm going to come back to what
13:55
code now I'm going to come back to what
13:55
code now I'm going to come back to what that means but before I do that I want
13:57
that means but before I do that I want
13:57
that means but before I do that I want to show you a few additional examples
13:59
to show you a few additional examples
13:59
to show you a few additional examples because I'm a big big big fan of
14:03
because I'm a big big big fan of
14:03
because I'm a big big big fan of visualizations because visualizations
14:05
visualizations because visualizations
14:05
visualizations because visualizations they kind of tap into this amazing power
14:07
they kind of tap into this amazing power
14:07
they kind of tap into this amazing power detector that we carry around in our
14:09
detector that we carry around in our
14:09
detector that we carry around in our head our visual brain right so by
14:13
head our visual brain right so by
14:13
head our visual brain right so by visualizing code we can
14:15
visualizing code we can
14:15
visualizing code we can within you know within a few seconds we
14:18
within you know within a few seconds we
14:18
within you know within a few seconds we can start to draw conclusions and
14:19
can start to draw conclusions and
14:19
can start to draw conclusions and compare various code bases and I want to
14:22
compare various code bases and I want to
14:22
compare various code bases and I want to show you an
14:24
show you an
14:24
show you an example so to the left here let's start
14:26
example so to the left here let's start
14:26
example so to the left here let's start with that one that's a visualization of
14:29
with that one that's a visualization of
14:29
with that one that's a visualization of the code quality in react from Facebook
14:32
the code quality in react from Facebook
14:32
the code quality in react from Facebook and react is a UI Library that's super
14:34
and react is a UI Library that's super
14:34
and react is a UI Library that's super popular these days most of you probably
14:37
popular these days most of you probably
14:37
popular these days most of you probably using it in one capacity or
14:39
using it in one capacity or
14:39
using it in one capacity or another and when I look at react I
14:41
another and when I look at react I
14:41
another and when I look at react I immediately see that the the react
14:43
immediately see that the the react
14:43
immediately see that the the react reconciler the part of the code that
14:45
reconciler the part of the code that
14:45
reconciler the part of the code that I've highlighted seems to have some code
14:48
I've highlighted seems to have some code
14:48
I've highlighted seems to have some code quality problems there's lots of yellow
14:51
quality problems there's lots of yellow
14:51
quality problems there's lots of yellow and red code inside it so if this is a
14:54
and red code inside it so if this is a
14:54
and red code inside it so if this is a code base that I'm responsible for then
14:57
code base that I'm responsible for then
14:57
code base that I'm responsible for then this is super important info to me
14:59
this is super important info to me
14:59
this is super important info to me because as we will soon see red and
15:01
because as we will soon see red and
15:01
because as we will soon see red and yellow code is going to constrain what
15:04
yellow code is going to constrain what
15:04
yellow code is going to constrain what we can do with the codebase and its road
15:06
we can do with the codebase and its road
15:06
we can do with the codebase and its road map going
15:07
map going
15:07
map going forward however I can contrast react to
15:10
forward however I can contrast react to
15:10
forward however I can contrast react to the codebase on the right and this is
15:13
the codebase on the right and this is
15:13
the codebase on the right and this is our Microsoft codebase this is Corr this
15:16
our Microsoft codebase this is Corr this
15:16
our Microsoft codebase this is Corr this is the runtime that you all use for net
15:20
is the runtime that you all use for net
15:20
is the runtime that you all use for net and what's interesting with the Corr is
15:22
and what's interesting with the Corr is
15:22
and what's interesting with the Corr is that it's a massive massive code base
15:25
that it's a massive massive code base
15:25
that it's a massive massive code base what you see on screen is more than
15:26
what you see on screen is more than
15:26
what you see on screen is more than eight and a half million lines of code
15:29
eight and a half million lines of code
15:29
eight and a half million lines of code so the reason I like this case study is
15:31
so the reason I like this case study is
15:31
so the reason I like this case study is because it kind of indicates that this
15:32
because it kind of indicates that this
15:32
because it kind of indicates that this visualization approach scales really
15:34
visualization approach scales really
15:34
visualization approach scales really well to large
15:36
well to large
15:36
well to large systems so what can we say about the
15:38
systems so what can we say about the
15:38
systems so what can we say about the code quality well we can immediately see
15:40
code quality well we can immediately see
15:41
code quality well we can immediately see that yeah all the automated tests in the
15:43
that yeah all the automated tests in the
15:43
that yeah all the automated tests in the bottom left corner seem to be very
15:45
bottom left corner seem to be very
15:45
bottom left corner seem to be very healthy right with very few
15:47
healthy right with very few
15:47
healthy right with very few exceptions but in the actual application
15:49
exceptions but in the actual application
15:49
exceptions but in the actual application code we seem to have some really
15:51
code we seem to have some really
15:51
code we seem to have some really complicated pieces of code you see this
15:53
complicated pieces of code you see this
15:53
complicated pieces of code you see this band of red code right it's like a Petr
15:55
band of red code right it's like a Petr
15:55
band of red code right it's like a Petr Dison mad almost right and the question
15:59
Dison mad almost right and the question
15:59
Dison mad almost right and the question is of course what do I do with this
16:00
is of course what do I do with this
16:00
is of course what do I do with this information how do I act upon this and
16:03
information how do I act upon this and
16:03
information how do I act upon this and that's something I want to cover towards
16:04
that's something I want to cover towards
16:04
that's something I want to cover towards the end of the
16:06
the end of the
16:06
the end of the session but before we do that we need to
16:09
session but before we do that we need to
16:09
session but before we do that we need to connect this to some kind of business
16:12
connect this to some kind of business
16:12
connect this to some kind of business objective because if we use a code
16:14
objective because if we use a code
16:14
objective because if we use a code quality score and we don't connect it to
16:16
quality score and we don't connect it to
16:16
quality score and we don't connect it to something that's meaningful to the
16:18
something that's meaningful to the
16:18
something that's meaningful to the business then code quality is always
16:20
business then code quality is always
16:20
business then code quality is always going to be a second priority right to
16:23
going to be a second priority right to
16:23
going to be a second priority right to really really make it the business
16:25
really really make it the business
16:25
really really make it the business Advantage it deserves to be we need to
16:27
Advantage it deserves to be we need to
16:27
Advantage it deserves to be we need to be able to explain gain and code quality
16:29
be able to explain gain and code quality
16:29
be able to explain gain and code quality to business
16:31
to business
16:31
to business people
16:32
people
16:32
people so two years ago me and my research
16:35
so two years ago me and my research
16:35
so two years ago me and my research colleague Dr Marcus Borg we set out to
16:38
colleague Dr Marcus Borg we set out to
16:38
colleague Dr Marcus Borg we set out to try to quantify the business impact of
16:40
try to quantify the business impact of
16:40
try to quantify the business impact of code
16:42
code
16:42
code quality and we choose to do this on
16:44
quality and we choose to do this on
16:44
quality and we choose to do this on Preparatory code basis because we wanted
16:47
Preparatory code basis because we wanted
16:47
Preparatory code basis because we wanted to make sure that this is data that's
16:48
to make sure that this is data that's
16:48
to make sure that this is data that's representative of what businesses work
16:51
representative of what businesses work
16:51
representative of what businesses work with and we did this by connecting to F
16:55
with and we did this by connecting to F
16:55
with and we did this by connecting to F nine companies that wanted to
16:57
nine companies that wanted to
16:57
nine companies that wanted to participate in the study
17:00
participate in the study
17:00
participate in the study each one of them had many many different
17:02
each one of them had many many different
17:02
each one of them had many many different Co products and code bases these are
17:04
Co products and code bases these are
17:04
Co products and code bases these are large
17:06
large
17:06
large Enterprises and we made sure to include
17:09
Enterprises and we made sure to include
17:09
Enterprises and we made sure to include companies from many different industry
17:11
companies from many different industry
17:11
companies from many different industry segments because we wanted to make sure
17:12
segments because we wanted to make sure
17:12
segments because we wanted to make sure that whatever we find is General to the
17:15
that whatever we find is General to the
17:15
that whatever we find is General to the software industry as a whole it's not
17:16
software industry as a whole it's not
17:17
software industry as a whole it's not something that's specific to one domain
17:19
something that's specific to one domain
17:19
something that's specific to one domain or one type of business and for the same
17:21
or one type of business and for the same
17:21
or one type of business and for the same reason we also made sure to include code
17:23
reason we also made sure to include code
17:23
reason we also made sure to include code in many different program languages
17:25
in many different program languages
17:25
in many different program languages because we wanted to make sure that
17:28
because we wanted to make sure that
17:28
because we wanted to make sure that whatever we find is something that
17:29
whatever we find is something that
17:29
whatever we find is something that applies no matter what program language
17:31
applies no matter what program language
17:31
applies no matter what program language you use because products today are
17:33
you use because products today are
17:33
you use because products today are typically built in a polyot fashion
17:36
typically built in a polyot fashion
17:36
typically built in a polyot fashion right so we had code written in C of
17:39
right so we had code written in C of
17:39
right so we had code written in C of course Java C++ JavaScript Python and so
17:43
course Java C++ JavaScript Python and so
17:43
course Java C++ JavaScript Python and so on and what I want to say before I show
17:46
on and what I want to say before I show
17:46
on and what I want to say before I show you the findings is that everything I
17:48
you the findings is that everything I
17:48
you the findings is that everything I show you is statistically significant
17:51
show you is statistically significant
17:51
show you is statistically significant meaning that it's very unlikely that our
17:53
meaning that it's very unlikely that our
17:53
meaning that it's very unlikely that our results are due to a fluke there's
17:55
results are due to a fluke there's
17:55
results are due to a fluke there's probably a real effect behind it and to
17:57
probably a real effect behind it and to
17:57
probably a real effect behind it and to guarantee that we had um our data
18:00
guarantee that we had um our data
18:00
guarantee that we had um our data published and peer reviewed and
18:02
published and peer reviewed and
18:02
published and peer reviewed and presented at academic
18:05
presented at academic
18:05
presented at academic conferences so what did we find is code
18:08
conferences so what did we find is code
18:08
conferences so what did we find is code quality important to a business well
18:11
quality important to a business well
18:11
quality important to a business well let's see I'm going to talk a bit about
18:13
let's see I'm going to talk a bit about
18:13
let's see I'm going to talk a bit about how we collected the data
18:16
how we collected the data
18:16
how we collected the data first so the first thing we did was to
18:20
first so the first thing we did was to
18:20
first so the first thing we did was to measure the time in development and we
18:22
measure the time in development and we
18:22
measure the time in development and we wanted to compare that across the code
18:25
wanted to compare that across the code
18:25
wanted to compare that across the code of categories so how do we know the time
18:27
of categories so how do we know the time
18:28
of categories so how do we know the time in development
18:29
in development
18:29
in development well code quality is easy we simply
18:31
well code quality is easy we simply
18:31
well code quality is easy we simply asked
18:33
asked
18:33
asked the companies that participated to run
18:36
the companies that participated to run
18:36
the companies that participated to run the tooling which gave us a code help
18:38
the tooling which gave us a code help
18:38
the tooling which gave us a code help score for every single file in their
18:40
score for every single file in their
18:40
score for every single file in their code bases right so we had a massive
18:41
code bases right so we had a massive
18:41
code bases right so we had a massive table with that but time in developments
18:44
table with that but time in developments
18:44
table with that but time in developments is more
18:45
is more
18:45
is more tricky what we didn't want to do was to
18:47
tricky what we didn't want to do was to
18:47
tricky what we didn't want to do was to force developers to self-report via some
18:50
force developers to self-report via some
18:50
force developers to self-report via some time spent field in J or something like
18:52
time spent field in J or something like
18:52
time spent field in J or something like that because if they would do that two
18:54
that because if they would do that two
18:54
that because if they would do that two things would happen first they would
18:56
things would happen first they would
18:56
things would happen first they would hate us for having them do that and that
18:59
hate us for having them do that and that
18:59
hate us for having them do that and that we didn't want second it would be
19:02
we didn't want second it would be
19:02
we didn't want second it would be inaccurate because they would basically
19:04
inaccurate because they would basically
19:04
inaccurate because they would basically just you know at the end of the week we
19:06
just you know at the end of the week we
19:06
just you know at the end of the week we put in whatever we think in those fields
19:08
put in whatever we think in those fields
19:09
put in whatever we think in those fields so we decided to automate it so we
19:11
so we decided to automate it so we
19:11
so we decided to automate it so we connected to tools like aure devops and
19:15
connected to tools like aure devops and
19:15
connected to tools like aure devops and jira and uh fetched information about
19:17
jira and uh fetched information about
19:17
jira and uh fetched information about the tickets when they started to work on
19:19
the tickets when they started to work on
19:19
the tickets when they started to work on them when they were completed and then
19:21
them when they were completed and then
19:21
them when they were completed and then we connected that to Version Control
19:22
we connected that to Version Control
19:22
we connected that to Version Control Data so we can trace it down the time
19:24
Data so we can trace it down the time
19:25
Data so we can trace it down the time spent at the file level so we basically
19:27
spent at the file level so we basically
19:27
spent at the file level so we basically calculated the cycle time for implenting
19:29
calculated the cycle time for implenting
19:29
calculated the cycle time for implenting features of fixing bag at the file level
19:32
features of fixing bag at the file level
19:32
features of fixing bag at the file level so we now had a massive table with
19:33
so we now had a massive table with
19:33
so we now had a massive table with unnown code of score and Unown time in
19:35
unnown code of score and Unown time in
19:35
unnown code of score and Unown time in development and that's what made it
19:37
development and that's what made it
19:37
development and that's what made it possible to calculate these
19:39
possible to calculate these
19:39
possible to calculate these statistics so what do they tell us well
19:42
statistics so what do they tell us well
19:42
statistics so what do they tell us well they tell us that if I have green code I
19:45
they tell us that if I have green code I
19:45
they tell us that if I have green code I can move more than twice as quick
19:47
can move more than twice as quick
19:47
can move more than twice as quick compared to someone that has red
19:49
compared to someone that has red
19:49
compared to someone that has red code is this important to a
19:52
code is this important to a
19:52
code is this important to a business well let's say that my company
19:56
business well let's say that my company
19:56
business well let's say that my company would have red code and uh let's say
19:59
would have red code and uh let's say
19:59
would have red code and uh let's say that we need to implement a new
20:01
that we need to implement a new
20:01
that we need to implement a new capability and it takes us two and a
20:02
capability and it takes us two and a
20:02
capability and it takes us two and a half months to do that a competitor with
20:05
half months to do that a competitor with
20:05
half months to do that a competitor with green code can get the same thing in
20:07
green code can get the same thing in
20:07
green code can get the same thing in less than a month it's going to be
20:08
less than a month it's going to be
20:08
less than a month it's going to be impossible to keep up so green code is
20:12
impossible to keep up so green code is
20:12
impossible to keep up so green code is very clearly a productivity benefit to a
20:14
very clearly a productivity benefit to a
20:14
very clearly a productivity benefit to a business you can get more things done
20:17
business you can get more things done
20:17
business you can get more things done than your competitors with poor
20:19
than your competitors with poor
20:19
than your competitors with poor codal however what we also noticed in
20:22
codal however what we also noticed in
20:22
codal however what we also noticed in that study we are well aware that
20:24
that study we are well aware that
20:24
that study we are well aware that averages can be a little bit misleading
20:26
averages can be a little bit misleading
20:26
averages can be a little bit misleading and the actual impact might be even
20:27
and the actual impact might be even
20:27
and the actual impact might be even larger
20:29
larger
20:29
larger so to study that we decided to look at
20:31
so to study that we decided to look at
20:31
so to study that we decided to look at the variation in task completion times
20:35
the variation in task completion times
20:35
the variation in task completion times and what we found here was very
20:38
and what we found here was very
20:38
and what we found here was very dramatic what we found was that the
20:40
dramatic what we found was that the
20:40
dramatic what we found was that the maximum time to implementing a ticket be
20:43
maximum time to implementing a ticket be
20:43
maximum time to implementing a ticket be it a new feature or bug fix could vary
20:47
it a new feature or bug fix could vary
20:47
it a new feature or bug fix could vary up to nine times longer 10 times longer
20:49
up to nine times longer 10 times longer
20:49
up to nine times longer 10 times longer for red code compared to Green code
20:52
for red code compared to Green code
20:52
for red code compared to Green code because when we look at the actual
20:54
because when we look at the actual
20:54
because when we look at the actual numbers behind this we see that the
20:55
numbers behind this we see that the
20:56
numbers behind this we see that the maximum time it takes to implement a
20:57
maximum time it takes to implement a
20:57
maximum time it takes to implement a ticket in green code is very very close
21:00
ticket in green code is very very close
21:00
ticket in green code is very very close to the average and that simply means
21:02
to the average and that simply means
21:02
to the average and that simply means that green code is predictable right it
21:05
that green code is predictable right it
21:05
that green code is predictable right it takes roughly the same time each time I
21:07
takes roughly the same time each time I
21:07
takes roughly the same time each time I work on it there are no nasty surprises
21:09
work on it there are no nasty surprises
21:09
work on it there are no nasty surprises in the
21:10
in the
21:10
in the code whereas I can see that for red code
21:13
code whereas I can see that for red code
21:13
code whereas I can see that for red code that relationship no longer holds
21:14
that relationship no longer holds
21:14
that relationship no longer holds through right it can take nine times
21:17
through right it can take nine times
21:17
through right it can take nine times longer is this relevant to a
21:20
longer is this relevant to a
21:20
longer is this relevant to a business I would like to think so
21:22
business I would like to think so
21:22
business I would like to think so because what this represents is
21:24
because what this represents is
21:24
because what this represents is uncertainty and if there's one thing
21:26
uncertainty and if there's one thing
21:26
uncertainty and if there's one thing I've learned through all these years in
21:28
I've learned through all these years in
21:28
I've learned through all these years in software it's that uncertainty is
21:29
software it's that uncertainty is
21:29
software it's that uncertainty is something that no one no one is
21:32
something that no one no one is
21:32
something that no one no one is comfortable with so imagine your
21:35
comfortable with so imagine your
21:35
comfortable with so imagine your business manager promising something to
21:37
business manager promising something to
21:37
business manager promising something to our key stakeholders maybe the company
21:40
our key stakeholders maybe the company
21:40
our key stakeholders maybe the company owners maybe to the board or maybe even
21:41
owners maybe to the board or maybe even
21:41
owners maybe to the board or maybe even to a customer that yeah we are building
21:44
to a customer that yeah we are building
21:44
to a customer that yeah we are building this feature now we're going to launch
21:46
this feature now we're going to launch
21:46
this feature now we're going to launch it in one
21:47
it in one
21:47
it in one month if they have red code they can
21:49
month if they have red code they can
21:49
month if they have red code they can make no such promises because as we've
21:52
make no such promises because as we've
21:52
make no such promises because as we've seen the data it can equally well take
21:54
seen the data it can equally well take
21:54
seen the data it can equally well take nine or 10 months and they going to look
21:56
nine or 10 months and they going to look
21:56
nine or 10 months and they going to look really really bad
21:58
really really bad
21:58
really really bad also if I put on my developer hat I'm
22:01
also if I put on my developer hat I'm
22:01
also if I put on my developer hat I'm also not a big fan of uncertainty
22:03
also not a big fan of uncertainty
22:03
also not a big fan of uncertainty because uncertainty is what's keeping me
22:05
because uncertainty is what's keeping me
22:05
because uncertainty is what's keeping me up late due to deadlines Mis deadlines
22:10
up late due to deadlines Mis deadlines
22:10
up late due to deadlines Mis deadlines overtime and stress so red code is
22:12
overtime and stress so red code is
22:12
overtime and stress so red code is disastrous on so many
22:15
disastrous on so many
22:15
disastrous on so many levels
22:17
levels
22:17
levels finally what we also did in this study
22:19
finally what we also did in this study
22:19
finally what we also did in this study was that we had as we had access to aure
22:22
was that we had as we had access to aure
22:22
was that we had as we had access to aure Dao and JRA data we could also figure
22:24
Dao and JRA data we could also figure
22:24
Dao and JRA data we could also figure out if the code was modified due to a
22:27
out if the code was modified due to a
22:27
out if the code was modified due to a new feature some re work or due to a bug
22:30
new feature some re work or due to a bug
22:30
new feature some re work or due to a bug fix and that made it possible to
22:32
fix and that made it possible to
22:32
fix and that made it possible to calculate defect
22:34
calculate defect
22:34
calculate defect densities so how what do you think how
22:37
densities so how what do you think how
22:37
densities so how what do you think how do you think defects uh differ across
22:40
do you think defects uh differ across
22:40
do you think defects uh differ across green yellow and red
22:43
green yellow and red
22:43
green yellow and red code well let's have a
22:46
code well let's have a
22:47
code well let's have a look and this is quite dramatic isn't it
22:49
look and this is quite dramatic isn't it
22:49
look and this is quite dramatic isn't it it turns out that red code has an
22:51
it turns out that red code has an
22:51
it turns out that red code has an average 15 times more defects than green
22:55
average 15 times more defects than green
22:55
average 15 times more defects than green code so let me ask the obvious question
22:58
code so let me ask the obvious question
22:58
code so let me ask the obvious question is this relevant to your
23:00
is this relevant to your
23:00
is this relevant to your business I definitely think so because
23:03
business I definitely think so because
23:03
business I definitely think so because if we having red code would ship 15
23:05
if we having red code would ship 15
23:06
if we having red code would ship 15 times as many defects as someone having
23:07
times as many defects as someone having
23:07
times as many defects as someone having green code then that would definitely
23:09
green code then that would definitely
23:09
green code then that would definitely impact the customer satisfaction and the
23:11
impact the customer satisfaction and the
23:11
impact the customer satisfaction and the product maturity
23:13
product maturity
23:13
product maturity experience and for us as developers red
23:15
experience and for us as developers red
23:15
experience and for us as developers red code is extremely
23:17
code is extremely
23:17
code is extremely stressful because all these defects that
23:19
stressful because all these defects that
23:19
stressful because all these defects that we ship into production they're going to
23:21
we ship into production they're going to
23:22
we ship into production they're going to come back in the shape of unplanned work
23:24
come back in the shape of unplanned work
23:24
come back in the shape of unplanned work where we have to contact switch
23:26
where we have to contact switch
23:26
where we have to contact switch right so red code is really really
23:30
right so red code is really really
23:30
right so red code is really really dangerous now if we have this data and
23:34
dangerous now if we have this data and
23:34
dangerous now if we have this data and visualizations like this how can we use
23:36
visualizations like this how can we use
23:36
visualizations like this how can we use them in our daily work well let me share
23:39
them in our daily work well let me share
23:39
them in our daily work well let me share the things I have seen work really
23:41
the things I have seen work really
23:41
the things I have seen work really really well and the things that I tend
23:42
really well and the things that I tend
23:42
really well and the things that I tend to do with this
23:44
to do with this
23:44
to do with this data the first thing that I think is
23:46
data the first thing that I think is
23:46
data the first thing that I think is super useful is that if we have a
23:48
super useful is that if we have a
23:48
super useful is that if we have a visualization like this we can use it to
23:50
visualization like this we can use it to
23:50
visualization like this we can use it to create situational awareness in our
23:52
create situational awareness in our
23:52
create situational awareness in our organization what I mean by that is that
23:54
organization what I mean by that is that
23:54
organization what I mean by that is that we can make sure that all stakeholders
23:57
we can make sure that all stakeholders
23:57
we can make sure that all stakeholders be it engineers developers or product
23:59
be it engineers developers or product
23:59
be it engineers developers or product people or managers we can share the same
24:02
people or managers we can share the same
24:02
people or managers we can share the same understanding the same view of where the
24:04
understanding the same view of where the
24:04
understanding the same view of where the strong and weak parts are in a code
24:06
strong and weak parts are in a code
24:06
strong and weak parts are in a code base and that's a really really good
24:09
base and that's a really really good
24:09
base and that's a really really good foundation because it helps us fight
24:12
foundation because it helps us fight
24:12
foundation because it helps us fight risk right because let's take an example
24:16
risk right because let's take an example
24:16
risk right because let's take an example here let's say that the codebase here to
24:18
here let's say that the codebase here to
24:18
here let's say that the codebase here to the right that I visualized let's say
24:20
the right that I visualized let's say
24:20
the right that I visualized let's say that it's our code base right so maybe
24:23
that it's our code base right so maybe
24:23
that it's our code base right so maybe we sit down in a Sprint planning meeting
24:25
we sit down in a Sprint planning meeting
24:25
we sit down in a Sprint planning meeting our product owner is attending and our
24:29
our product owner is attending and our
24:29
our product owner is attending and our product owner tells us that she has
24:31
product owner tells us that she has
24:31
product owner tells us that she has planned five major features in the react
24:33
planned five major features in the react
24:33
planned five major features in the react reconciler
24:35
reconciler
24:35
reconciler part now given what we know about red
24:38
part now given what we know about red
24:38
part now given what we know about red and yellow code we know that this is a
24:40
and yellow code we know that this is a
24:40
and yellow code we know that this is a massive risk that's a large risk that
24:42
massive risk that's a large risk that
24:42
massive risk that's a large risk that we're going to um need much more time
24:44
we're going to um need much more time
24:44
we're going to um need much more time than we plan for and that we are going
24:46
than we plan for and that we are going
24:46
than we plan for and that we are going to deliver with poor quality many
24:48
to deliver with poor quality many
24:48
to deliver with poor quality many defects so using this data we can have a
24:51
defects so using this data we can have a
24:51
defects so using this data we can have a conversation with it makes sense to
24:53
conversation with it makes sense to
24:53
conversation with it makes sense to First refactor the code so that we can
24:55
First refactor the code so that we can
24:55
First refactor the code so that we can safely Implement all these features and
24:58
safely Implement all these features and
24:58
safely Implement all these features and that brings me to the last use case my
25:00
that brings me to the last use case my
25:00
that brings me to the last use case my personal favorite and perhaps what I
25:02
personal favorite and perhaps what I
25:02
personal favorite and perhaps what I like the most about this type of data is
25:04
like the most about this type of data is
25:04
like the most about this type of data is that it makes it possible to finally
25:06
that it makes it possible to finally
25:06
that it makes it possible to finally build a business case for improvements
25:08
build a business case for improvements
25:08
build a business case for improvements and large scale refactorings because
25:11
and large scale refactorings because
25:11
and large scale refactorings because these can now come with a business
25:13
these can now come with a business
25:13
these can now come with a business expectation So based on the data we can
25:17
expectation So based on the data we can
25:17
expectation So based on the data we can we can give guarantees that if we have
25:19
we can give guarantees that if we have
25:19
we can give guarantees that if we have code like in the react reconciler if we
25:21
code like in the react reconciler if we
25:21
code like in the react reconciler if we take that red code refactor it and make
25:23
take that red code refactor it and make
25:23
take that red code refactor it and make it healthy then everything that we have
25:26
it healthy then everything that we have
25:26
it healthy then everything that we have planned in the road mapap for that part
25:28
planned in the road mapap for that part
25:28
planned in the road mapap for that part particular component we can now say that
25:30
particular component we can now say that
25:30
particular component we can now say that on average we're going to be twice as
25:31
on average we're going to be twice as
25:31
on average we're going to be twice as quick with implementing those features
25:33
quick with implementing those features
25:33
quick with implementing those features and we going significantly reduce the
25:35
and we going significantly reduce the
25:35
and we going significantly reduce the risk we're going to have 15 times fewer
25:37
risk we're going to have 15 times fewer
25:37
risk we're going to have 15 times fewer defects on average and that's a very
25:39
defects on average and that's a very
25:39
defects on average and that's a very very quantifiable
25:43
win now I still have a couple of minutes
25:46
win now I still have a couple of minutes
25:46
win now I still have a couple of minutes before uh taking questions so I wanted
25:49
before uh taking questions so I wanted
25:49
before uh taking questions so I wanted to take on one more related
25:52
to take on one more related
25:52
to take on one more related challenge how do we prioritize
25:55
challenge how do we prioritize
25:55
challenge how do we prioritize remediations to large amounts of red
25:57
remediations to large amounts of red
25:57
remediations to large amounts of red code
25:59
code
25:59
code so let's travel back to coreclr as an
26:03
so let's travel back to coreclr as an
26:03
so let's travel back to coreclr as an example 8.5 million lines of code in
26:06
example 8.5 million lines of code in
26:06
example 8.5 million lines of code in total if we just sum up the red code we
26:08
total if we just sum up the red code we
26:08
total if we just sum up the red code we probably end up with two two and a half
26:10
probably end up with two two and a half
26:10
probably end up with two two and a half million lines of code how long would it
26:13
million lines of code how long would it
26:13
million lines of code how long would it take us to refactor two and a half
26:15
take us to refactor two and a half
26:15
take us to refactor two and a half million lines of
26:16
million lines of
26:16
million lines of code five years 10 years I don't know
26:20
code five years 10 years I don't know
26:21
code five years 10 years I don't know but it's very clear that we would be
26:22
but it's very clear that we would be
26:22
but it's very clear that we would be long out of business before we even get
26:24
long out of business before we even get
26:24
long out of business before we even get there right because as a business we
26:26
there right because as a business we
26:26
there right because as a business we always need to balance improvements to
26:29
always need to balance improvements to
26:29
always need to balance improvements to existing code versus adding new features
26:32
existing code versus adding new features
26:32
existing code versus adding new features right it's a continuously changing
26:35
right it's a continuously changing
26:35
right it's a continuously changing space so how can we do this well this is
26:39
space so how can we do this well this is
26:39
space so how can we do this well this is where our set of techniques known as
26:41
where our set of techniques known as
26:41
where our set of techniques known as behavioral code analysis can
26:44
behavioral code analysis can
26:44
behavioral code analysis can help so for those of you who aren't
26:47
help so for those of you who aren't
26:47
help so for those of you who aren't familiar with behavioral code analysis
26:49
familiar with behavioral code analysis
26:49
familiar with behavioral code analysis the key idea here is that in our
26:51
the key idea here is that in our
26:51
the key idea here is that in our behavioral code analysis the code is
26:54
behavioral code analysis the code is
26:54
behavioral code analysis the code is important it's really really important
26:57
important it's really really important
26:57
important it's really really important to know if the code is healthy or not
26:59
to know if the code is healthy or not
26:59
to know if the code is healthy or not right so that's like the
27:01
right so that's like the
27:01
right so that's like the foundation once we have that it's even
27:04
foundation once we have that it's even
27:04
foundation once we have that it's even more important to understand how we as
27:06
more important to understand how we as
27:06
more important to understand how we as developers interact with the system
27:08
developers interact with the system
27:08
developers interact with the system we're building that's the behavior part
27:11
we're building that's the behavior part
27:11
we're building that's the behavior part of it so it's an
27:13
of it so it's an
27:13
of it so it's an intersection between code and
27:15
intersection between code and
27:15
intersection between code and people and to do a behavioral code
27:17
people and to do a behavioral code
27:18
people and to do a behavioral code analysis we obviously need some
27:19
analysis we obviously need some
27:19
analysis we obviously need some behavioral data on how we as developers
27:22
behavioral data on how we as developers
27:22
behavioral data on how we as developers work with the code how can we possibly
27:25
work with the code how can we possibly
27:25
work with the code how can we possibly get such a
27:26
get such a
27:26
get such a thing the good news are you all already
27:30
thing the good news are you all already
27:30
thing the good news are you all already have all the data you need we might just
27:32
have all the data you need we might just
27:32
have all the data you need we might just not be used to think about it that way
27:35
not be used to think about it that way
27:35
not be used to think about it that way what I'm referring to is Version Control
27:38
what I'm referring to is Version Control
27:38
what I'm referring to is Version Control Data Git it turns out that git is
27:42
Data Git it turns out that git is
27:42
Data Git it turns out that git is something we have used for years more or
27:44
something we have used for years more or
27:45
something we have used for years more or less as an overly complicated backup
27:46
less as an overly complicated backup
27:46
less as an overly complicated backup system occasionally maybe as a
27:48
system occasionally maybe as a
27:48
system occasionally maybe as a collaboration tool but when doing so we
27:51
collaboration tool but when doing so we
27:51
collaboration tool but when doing so we have built up this wonderful data source
27:53
have built up this wonderful data source
27:53
have built up this wonderful data source over how we as Engineers have interacted
27:56
over how we as Engineers have interacted
27:56
over how we as Engineers have interacted with the code we're building and there
27:58
with the code we're building and there
27:58
with the code we're building and there are many many analysis we can do from
28:00
are many many analysis we can do from
28:00
are many many analysis we can do from git I write about lots of them in my
28:03
git I write about lots of them in my
28:03
git I write about lots of them in my book but one of my favorites when
28:06
book but one of my favorites when
28:06
book but one of my favorites when particular when it comes to technical de
28:08
particular when it comes to technical de
28:08
particular when it comes to technical de is that git data gives us a Time
28:11
is that git data gives us a Time
28:11
is that git data gives us a Time Dimension into our code we can basically
28:14
Dimension into our code we can basically
28:14
Dimension into our code we can basically tell the story of how the system evolved
28:17
tell the story of how the system evolved
28:17
tell the story of how the system evolved where we worked how frequently we worked
28:19
where we worked how frequently we worked
28:19
where we worked how frequently we worked in various parts and what happened over
28:22
in various parts and what happened over
28:22
in various parts and what happened over time and this is something we can use to
28:25
time and this is something we can use to
28:25
time and this is something we can use to prioritize technical do
28:28
prioritize technical do
28:28
prioritize technical do so instead of showing you a video
28:29
so instead of showing you a video
28:29
so instead of showing you a video recording I wanted to do a quick demo on
28:32
recording I wanted to do a quick demo on
28:32
recording I wanted to do a quick demo on how it can look on a real code base and
28:34
how it can look on a real code base and
28:34
how it can look on a real code base and then I'm going to wrap up the
28:35
then I'm going to wrap up the
28:35
then I'm going to wrap up the presentation and take questions after
28:39
presentation and take questions after
28:39
presentation and take questions after that so let's see it's always
28:41
that so let's see it's always
28:41
that so let's see it's always interesting to do a live
28:43
interesting to do a live
28:43
interesting to do a live demo uh you should be able to see my web
28:46
demo uh you should be able to see my web
28:46
demo uh you should be able to see my web browser
28:47
browser
28:47
browser now and the system I'm analyzing is
28:50
now and the system I'm analyzing is
28:50
now and the system I'm analyzing is react from Facebook that we looked at
28:53
react from Facebook that we looked at
28:53
react from Facebook that we looked at before and we can easily see that there
28:55
before and we can easily see that there
28:55
before and we can easily see that there are parts of the code that are unhealthy
28:59
are parts of the code that are unhealthy
28:59
are parts of the code that are unhealthy I see like two modules right there's a
29:01
I see like two modules right there's a
29:01
I see like two modules right there's a pattern here so to the left I have
29:04
pattern here so to the left I have
29:04
pattern here so to the left I have something called benchmarks where there
29:05
something called benchmarks where there
29:05
something called benchmarks where there is plenty of unhealthy code and to the
29:08
is plenty of unhealthy code and to the
29:08
is plenty of unhealthy code and to the right I have the react reconciler with
29:10
right I have the react reconciler with
29:10
right I have the react reconciler with unhealthy code the question is where do
29:13
unhealthy code the question is where do
29:13
unhealthy code the question is where do I start if I want to pay down technical
29:15
I start if I want to pay down technical
29:15
I start if I want to pay down technical debt and
29:17
debt and
29:17
debt and improve the interesting thing here is
29:19
improve the interesting thing here is
29:19
improve the interesting thing here is that if we look at Version Control Data
29:21
that if we look at Version Control Data
29:21
that if we look at Version Control Data we can visualize that using a concept
29:23
we can visualize that using a concept
29:23
we can visualize that using a concept called hotspots and I'm going to explain
29:25
called hotspots and I'm going to explain
29:25
called hotspots and I'm going to explain what that is so if I click here on
29:27
what that is so if I click here on
29:27
what that is so if I click here on hotspots what I do now is that I go to
29:29
hotspots what I do now is that I go to
29:29
hotspots what I do now is that I go to verse control and I simply calculate how
29:31
verse control and I simply calculate how
29:31
verse control and I simply calculate how many commits have we done to every
29:33
many commits have we done to every
29:33
many commits have we done to every single piece of code here over the past
29:35
single piece of code here over the past
29:35
single piece of code here over the past year and what I see now is that the more
29:38
year and what I see now is that the more
29:38
year and what I see now is that the more red it is the higher the development
29:40
red it is the higher the development
29:40
red it is the higher the development frequency right the more commits we have
29:42
frequency right the more commits we have
29:42
frequency right the more commits we have done in that part of the code the more
29:43
done in that part of the code the more
29:43
done in that part of the code the more relevant the code is and what I see now
29:46
relevant the code is and what I see now
29:46
relevant the code is and what I see now might be surprising I see that this
29:48
might be surprising I see that this
29:48
might be surprising I see that this benchmarks package where I had this red
29:51
benchmarks package where I had this red
29:51
benchmarks package where I had this red code is very stable we haven't worked on
29:54
code is very stable we haven't worked on
29:54
code is very stable we haven't worked on it for years so starting to fix some
29:57
it for years so starting to fix some
29:57
it for years so starting to fix some code down there there won't have any big
29:59
code down there there won't have any big
29:59
code down there there won't have any big payoff right however I see that there's
30:02
payoff right however I see that there's
30:02
payoff right however I see that there's a lot of development activity in the
30:04
a lot of development activity in the
30:04
a lot of development activity in the react reconciler so I would definitely
30:06
react reconciler so I would definitely
30:06
react reconciler so I would definitely zoom in here and see if these hotspots
30:09
zoom in here and see if these hotspots
30:09
zoom in here and see if these hotspots are unhealthy so I let me click on a
30:11
are unhealthy so I let me click on a
30:11
are unhealthy so I let me click on a hotspot and I immediately see that this
30:14
hotspot and I immediately see that this
30:14
hotspot and I immediately see that this is indeed unhealthy code so what I have
30:17
is indeed unhealthy code so what I have
30:17
is indeed unhealthy code so what I have identified now is code that lacks in
30:20
identified now is code that lacks in
30:20
identified now is code that lacks in quality poor quality code that is also
30:22
quality poor quality code that is also
30:22
quality poor quality code that is also worked on all the time and that's like
30:25
worked on all the time and that's like
30:25
worked on all the time and that's like the definition of technical depth with a
30:27
the definition of technical depth with a
30:27
the definition of technical depth with a high interest
30:29
high interest
30:29
high interest rate so I want to show you how Behavior
30:33
rate so I want to show you how Behavior
30:33
rate so I want to show you how Behavior code analysis techniques like a hotspot
30:35
code analysis techniques like a hotspot
30:35
code analysis techniques like a hotspot can help you separate the technical
30:37
can help you separate the technical
30:37
can help you separate the technical depth that you have to fix from the
30:39
depth that you have to fix from the
30:39
depth that you have to fix from the technical depth you can live with and I
30:41
technical depth you can live with and I
30:41
technical depth you can live with and I want to clarify that with the slide
30:43
want to clarify that with the slide
30:43
want to clarify that with the slide because it's so
30:45
because it's so
30:45
because it's so important so technical depth has two
30:47
important so technical depth has two
30:47
important so technical depth has two components one is the actual loan you
30:50
components one is the actual loan you
30:50
components one is the actual loan you have taken on and this is something you
30:52
have taken on and this is something you
30:52
have taken on and this is something you can measure by uh code quality right red
30:56
can measure by uh code quality right red
30:56
can measure by uh code quality right red or green code but the other is the
30:59
or green code but the other is the
30:59
or green code but the other is the interest and if we look at this
31:01
interest and if we look at this
31:01
interest and if we look at this visualization I have a slightly
31:03
visualization I have a slightly
31:03
visualization I have a slightly different view of this code base so on
31:05
different view of this code base so on
31:05
different view of this code base so on the xaxis you have each file with source
31:08
the xaxis you have each file with source
31:08
the xaxis you have each file with source code and they are sorted according to
31:10
code and they are sorted according to
31:10
code and they are sorted according to the ex change frequency that is how many
31:12
the ex change frequency that is how many
31:12
the ex change frequency that is how many commits have I done to each piece of
31:13
commits have I done to each piece of
31:13
commits have I done to each piece of code that's what I see on the y-axis and
31:16
code that's what I see on the y-axis and
31:16
code that's what I see on the y-axis and if I look at this visualization I see
31:18
if I look at this visualization I see
31:18
if I look at this visualization I see that it forms a power law
31:20
that it forms a power law
31:20
that it forms a power law shape and this is something I've seen in
31:22
shape and this is something I've seen in
31:22
shape and this is something I've seen in every single codebase I've ever analyzed
31:25
every single codebase I've ever analyzed
31:25
every single codebase I've ever analyzed and I probably analyzed three 400
31:27
and I probably analyzed three 400
31:27
and I probably analyzed three 400 codebases by now so this seems to be the
31:29
codebases by now so this seems to be the
31:29
codebases by now so this seems to be the way software evolves and this is
31:32
way software evolves and this is
31:32
way software evolves and this is important to us and it's actually good
31:34
important to us and it's actually good
31:34
important to us and it's actually good news because what this means is that
31:37
news because what this means is that
31:37
news because what this means is that most of our code is going to be in the
31:38
most of our code is going to be in the
31:38
most of our code is going to be in the long tail so this is code that's rarely
31:41
long tail so this is code that's rarely
31:41
long tail so this is code that's rarely ever touched and if we have some yellow
31:43
ever touched and if we have some yellow
31:43
ever touched and if we have some yellow or red code down there then that's
31:45
or red code down there then that's
31:45
or red code down there then that's technical depth we want to be aware of
31:47
technical depth we want to be aware of
31:47
technical depth we want to be aware of it but we can probably live with it
31:49
it but we can probably live with it
31:49
it but we can probably live with it because the interest is so low on the
31:52
because the interest is so low on the
31:52
because the interest is so low on the other hand we see that most development
31:54
other hand we see that most development
31:54
other hand we see that most development activity is at the head of this Power
31:56
activity is at the head of this Power
31:56
activity is at the head of this Power log curve so that means that in a
32:00
log curve so that means that in a
32:00
log curve so that means that in a Hotpot we cannot afford any technical
32:03
Hotpot we cannot afford any technical
32:03
Hotpot we cannot afford any technical debt at all that's the code we really
32:05
debt at all that's the code we really
32:05
debt at all that's the code we really really need to keep green because if we
32:07
really need to keep green because if we
32:07
really need to keep green because if we fail to do so the costs of that
32:09
fail to do so the costs of that
32:09
fail to do so the costs of that technical depth are going to explode due
32:11
technical depth are going to explode due
32:11
technical depth are going to explode due to the high development frequency of the
32:13
to the high development frequency of the
32:13
to the high development frequency of the code right it's a very high interest on
32:15
code right it's a very high interest on
32:15
code right it's a very high interest on that technical
32:16
that technical
32:16
that technical depth and what I see in our research is
32:20
depth and what I see in our research is
32:20
depth and what I see in our research is that the hotspots tend to make up a very
32:22
that the hotspots tend to make up a very
32:22
that the hotspots tend to make up a very small part of the code base maybe just
32:24
small part of the code base maybe just
32:24
small part of the code base maybe just two three
32:26
two three
32:26
two three four% but attract on unproportionally
32:29
four% but attract on unproportionally
32:29
four% but attract on unproportionally large part of the development activity
32:32
large part of the development activity
32:32
large part of the development activity so starting to pay down technical debt
32:34
so starting to pay down technical debt
32:34
so starting to pay down technical debt in a hotspot is likely to give you a
32:36
in a hotspot is likely to give you a
32:36
in a hotspot is likely to give you a very big short-term benefit as
32:39
very big short-term benefit as
32:40
very big short-term benefit as well and with that I've come to the end
32:42
well and with that I've come to the end
32:42
well and with that I've come to the end of my presentation what I want to show
32:44
of my presentation what I want to show
32:44
of my presentation what I want to show you today is that to really really uh
32:46
you today is that to really really uh
32:46
you today is that to really really uh make code quality or competitive
32:48
make code quality or competitive
32:48
make code quality or competitive Advantage you need to have two different
32:50
Advantage you need to have two different
32:50
Advantage you need to have two different dimensions one is the quality Dimension
32:53
dimensions one is the quality Dimension
32:53
dimensions one is the quality Dimension that you can measure with code Health
32:54
that you can measure with code Health
32:54
that you can measure with code Health where you see if the code is good or bad
32:57
where you see if the code is good or bad
32:57
where you see if the code is good or bad and the other is the relevance Dimension
32:59
and the other is the relevance Dimension
32:59
and the other is the relevance Dimension that you can measure with hotspots so
33:01
that you can measure with hotspots so
33:01
that you can measure with hotspots so that is how you know the impact and
33:03
that is how you know the impact and
33:03
that is how you know the impact and priorities of any cool quality
33:06
priorities of any cool quality
33:06
priorities of any cool quality findings and I started out saying that
33:09
findings and I started out saying that
33:09
findings and I started out saying that um there seems to be this conception in
33:11
um there seems to be this conception in
33:12
um there seems to be this conception in Industry that there is a trade-off
33:13
Industry that there is a trade-off
33:13
Industry that there is a trade-off between speed and quality but what we
33:15
between speed and quality but what we
33:15
between speed and quality but what we see in our research in our data is that
33:18
see in our research in our data is that
33:18
see in our research in our data is that the opposite seems to be true and I like
33:20
the opposite seems to be true and I like
33:20
the opposite seems to be true and I like to think that with 15 times fewer
33:22
to think that with 15 times fewer
33:22
to think that with 15 times fewer defects twice the development speed and
33:25
defects twice the development speed and
33:25
defects twice the development speed and nine times lower onc in completion times
33:27
nine times lower onc in completion times
33:27
nine times lower onc in completion times for talk asks the business advantage of
33:29
for talk asks the business advantage of
33:29
for talk asks the business advantage of cod quality should be really really
33:33
cod quality should be really really
33:33
cod quality should be really really clear and if you want to know more about
33:35
clear and if you want to know more about
33:35
clear and if you want to know more about this then I have a number of resources
33:37
this then I have a number of resources
33:37
this then I have a number of resources for you here uh the links will take you
33:40
for you here uh the links will take you
33:40
for you here uh the links will take you to a white paper that summarizes uh the
33:43
to a white paper that summarizes uh the
33:43
to a white paper that summarizes uh the key uh code Health findings that I
33:45
key uh code Health findings that I
33:45
key uh code Health findings that I talked about the business impact there's
33:48
talked about the business impact there's
33:48
talked about the business impact there's also an actual research paper behind it
33:50
also an actual research paper behind it
33:50
also an actual research paper behind it so if you want to go into all the
33:51
so if you want to go into all the
33:51
so if you want to go into all the details and figure out that hey how did
33:53
details and figure out that hey how did
33:53
details and figure out that hey how did they measure this or how did they
33:54
they measure this or how did they
33:55
they measure this or how did they control for this bias then the research
33:57
control for this bias then the research
33:57
control for this bias then the research paper is your first
33:58
paper is your first
33:58
paper is your first brand and uh to do the actual research
34:02
brand and uh to do the actual research
34:02
brand and uh to do the actual research and to visualize it I'm using the codin
34:04
and to visualize it I'm using the codin
34:04
and to visualize it I'm using the codin Tool uh where I work so if you're
34:07
Tool uh where I work so if you're
34:07
Tool uh where I work so if you're interested in trying this out on your
34:09
interested in trying this out on your
34:09
interested in trying this out on your own codebase then I highly encourage you
34:11
own codebase then I highly encourage you
34:11
own codebase then I highly encourage you to give it a shot I think you will find
34:12
to give it a shot I think you will find
34:12
to give it a shot I think you will find it interesting to see what your code
34:14
it interesting to see what your code
34:14
it interesting to see what your code looks like and you can go to codes.com
34:16
looks like and you can go to codes.com
34:16
looks like and you can go to codes.com and just test it out for free and
34:19
and just test it out for free and
34:19
and just test it out for free and finally if you want to dive deeper into
34:21
finally if you want to dive deeper into
34:21
finally if you want to dive deeper into Behavior code analysis then you could
34:23
Behavior code analysis then you could
34:23
Behavior code analysis then you could probably do worse than reading the
34:24
probably do worse than reading the
34:24
probably do worse than reading the second edition of your codess crime
34:26
second edition of your codess crime
34:26
second edition of your codess crime scene it's a brand new second edition
34:29
scene it's a brand new second edition
34:29
scene it's a brand new second edition that goes into Cod quality hotspots and
34:31
that goes into Cod quality hotspots and
34:31
that goes into Cod quality hotspots and many many other things around successful
34:34
many many other things around successful
34:34
many many other things around successful software with that said I'd like to
34:36
software with that said I'd like to
34:36
software with that said I'd like to thank you a lot for listening to me and
34:38
thank you a lot for listening to me and
34:38
thank you a lot for listening to me and I'm very much looking forward to the
34:40
I'm very much looking forward to the
34:40
I'm very much looking forward to the questions so thank you very much
34:42
questions so thank you very much
34:42
questions so thank you very much [Music]