Project Description

Mentors

Tasks completed

New imports included in ChemoSpec

Example plots with `patchwork`,`ggrepel` and `ggplot2` packages

Package patchwork was used for this plot as different geoms were needed in the top two panels, which cannot be handled by ggplot2. However, patchwork is not supported by plotly so this function was not enhanced by plotly.

This is one of several plots in which the labeling of points was greatly improved compared to the base graphics implementation. In particular, using ggrepel gives a much better visual experience with less code on our end.

What’s done!

After several weeks of experimentation, it was decided that creating new graphics modes using ggplot2 and plotly would greatly enhance ChemoSpec, and that extension to shiny would not add much to user experience. With that change in plans by the mentors, all project goals were reached.
New function chkGraphicsOpt has been created to check the graphics mode.
ggplot2 graphics mode has been added to the functions listed in Phase 1 tasklist.
plotly graphics mode has been added to the functions listed in Phase 2 tasklist.
Many helper functions such as .prepLegendCoords and .ggAddLegend were written to avoid repeating code, and these will also make it much easier to convert other functions to ggplot2 mode in the future.

What’s next for the project?!

Some functions that were not in the original plan could still benefit from adding the new graphics mode.
plotLoadings was created using the patchwork package for the best scientific and visual experience.patchwork does not yet work with plotly, so in the future if this becomes possible then plotly should be integrated.
annotate_custom() function is implemented in ggplot2 but it is yet to be implemented in plotly mode so a couple of labels might not work in plotly graphics mode. As soon as annotate_custom() is translated by the plotly developers, ChemoSpec will be able to use to use it in plotly graphics mode as well.
The project mentors will be double-checking everything and preparing both ChemoSpec and ChemoSpecUtils for release within a month.

Weekly Blogs with Commits :

Community Bonding period

Community bonding period started right after the announcement of results. The very first task was to set up a solid communication medium as there is a difference of about 12 hour time difference. Erick set up the slack channel for the project. Most of my time during this period went in understanding the codebase of ChemoSpec and ChemoSpecUtils. Along with learning about the working of the package I also learned about github actions, build and check process of the package, roxygen documentation and tinytest. Bryan setup the new branch gsoc1 and the mechanism for switching between the graphic modes of the plotting function.The graphics option can now be switched by simply options(chemoSpecGraphics= "any") where any is the name of the graphic mode. After all the learning I wrote my first function chkGrafOpt() which I later renamed to chkGraphicsOpt() for better understanding of the function. It is a very simple function but yet very important. Even though it was very simple to code I still learnt something new from it. To test the function I had to create the test for it. As ChemoSpec is majorly used for plotting the spectra, it uses tinytest for testing. After the successful completion and documentation of the function, my first pull request was merged. It felt great! After that my second pull request also got merged.

My commits during this period :

https://github.com/bryanhanson/ChemoSpec/pull/33/commits
https://github.com/bryanhanson/ChemoSpec/pull/37/commits

Week 1

With all the preparations done for the project, I started my work on the plotSpectra function. The plotSpectra plots the spectra object. Unlike the base graphics, I had to come up with a new way for plotting data in ggplot2 as things work slightly different in ggplot2. I used melt function from reshape2 to organize the data which helped me in plotting multiple geom_lines. As ggplot2 was using NPC(Normalized Plot Coordinates) to plot the legend on the plot, we decided to implement the same in base graphics too. It was a little trickier to implement in ggplot2 and so I had to come up with something new. I used grobTree and textgrob for capturing the group names along with their group colors and used annotate_custom to attach it with the ggplot2 plot. Apart from build and check the function is working fine and hopefully I will be able to merge it without any error the following week.

As this was my first week I made a lot mistakes but Bryan was always there to the rescue. It was an amazing week and I learnt a lot of cool stuff. Let’s go to Week 2!

My commits during this period :

https://github.com/bryanhanson/ChemoSpec/pull/38/commits

Week 2

After an amazing first week, things have started to pick up the pace. This week started with me fixing the minor bugs in the plotSpectra function. After plotSpectra function was merged, I decided to work on the loopThruSpectra function. For loopThruSpectra function we decided to make some changes in the ggplot2 mode. In base grahpics mode the user goes through all the spectra available in the data one by one but in ggplot2 mode the user will be able to review all the spectra stored as all of them will be stacked vertically in a single plot. Because loopThruSpectra function will basically help in reviewing all the spectra stored in the data, Bryan decided to rename the function as reviewAllSpectra. After the completion of reviewAllSpectra function and it’s documentation I decided to work on surveySpectra function. The main thing in surveySpectra function was to implement the faceting. Once I understood the concept of faceting in ggplot2, I completed the function and it’s documentaion without any problem. During this week Bryan also made some changes in the base graphics mode to speed up the function. After this I decided to work on the plotLoading function rather than surveySpectra2 function as it seemed a little less complicated. plotLoading was kind of similar to surveySpectra function so I was able to wrap it up quickly without any problems. I first used geom_area for the plot, but later on close inspection of the function in base mode on metMUD1 dataset I decided to switch to geom_segment.

Overall this was great week with lot’s of new learning! On to Week 3!

My commits during this period :

https://github.com/bryanhanson/ChemoSpec/pull/39/commits
https://github.com/bryanhanson/ChemoSpec/pull/40/commits
https://github.com/bryanhanson/ChemoSpec/pull/41/commits
https://github.com/bryanhanson/ChemoSpec/pull/42/commits

Week 3

This week started with me working on the ggplot2 mode for surveySpectra2 function. This function was a lot similar to the plotSpectra function so after understanding the base mode I was able to figure out the logic for the same for ggplot2 mode. Now that most of the 1d plot functions were over, it was time for 2d plot functions. Working on the even the first function turned out to be a very big deal. Most of the 2d functions in ChemoSpec depends on the ChemoSpecUtils package. I decided to start my work with pcaDiag function. pcaDiag was fairly simple to code until I got to the .labelExtreme part. .labelExtreme is a utility function from ChemoSpecUtils which plots the label of the extreme data points. As the base mode is slightly different from ggplot2 mode, we decided to make another function getExtremeCoords specifically for ggplot2 mode. getExtremeCoords uses labelExtreme function to get the coordinates of the labels and returns a list of them. The main idea for creating this function is to get the data and not interfere with base mode code. Finally, I have completed the code for pcaDiag function. Some problems still remains regarding build and check though but these will be rectified in the next week.

On to week 4!

My commits during this period :

https://github.com/bryanhanson/ChemoSpec/pull/43/commits
https://github.com/bryanhanson/ChemoSpecUtils/pull/2/commits
https://github.com/bryanhanson/ChemoSpecUtils/pull/3/commits

Week 4

The very first task in this week was to solve the build and check process issue. Because ChemoSpec depends on ChemoSpecUtils we have decided to only locally run the build and check process for now. Since pcaDiag function was working fine locally, we merged it. After pcaDiag I decided to work on splotSpectra. splotSpectra was relativly simple to code and so I was able to do it without any problem. After this I worked on plotSpectraDist function. This function was also relativly simple to code, however I faced a problem in labeling part. This wasn’t the first time I was facing this problem, so for now I decided to use a fudge factor to make the labels seperated from the plot points. While working on all the new functions we came across a new package patchwork. I had already worked on plotLoading function’s ggplot2 mode but we thought of giving a try to patchwork package.patchwork package helps user in organizing the different plots in a grid. The plot has turned out to be really nice, maybe even better than before. However, we are still thinking about the function. Overall it required quite a lot of thinking as I had to change the way data was organized in the base graphics for the code to work. Best part of coding this function was discovering a new way to plot multiple ggplots using for loop. For now we are sticking to plot made using patchwork.

Next week we have use R! conference!

My commits during this period :

https://github.com/bryanhanson/ChemoSpec/pull/44/commits
https://github.com/bryanhanson/ChemoSpec/pull/45/commits
https://github.com/bryanhanson/ChemoSpec/pull/47/commits
https://github.com/bryanhanson/ChemoSpec/pull/49/commits

Week 5

With this week, the first half of GSoC has come to an end. This week was different from all the previous weeks. This week I attended use R! conference. It was a really nice experience for me as I got to know more about the R language,it’s packages and it’s community.

Apart form use R! conference I worked on the remaining funcitons. I started my work with plot2Loadings function. As plot2loading was similar to functions previously done by me, I was able to write the ggplot2 mode without any problem. While I was completing the function, Bryan came across a bug in surveySpectra function. So after I completed the plot2loading function, I worked on the surveySpectra function and added the missing code. After completing these functions I started my work on plotScree function in the ChemoSpecUtils package. The print statement was causing a little problem in rendering of the ggplots, so we decided to “return” the ggplots rather than printing them. This will ensure that ggplot plot always gets printed after the function is called. Lastly, I completed the code for ggplot2 mode of the plotScree function.

My commits during this period :

https://github.com/bryanhanson/ChemoSpec/pull/50/commits
https://github.com/bryanhanson/ChemoSpec/pull/51/commits
https://github.com/bryanhanson/ChemoSpec/pull/52/commits
https://github.com/bryanhanson/ChemoSpecUtils/pull/4/commits

Week 6

This week was probably the most toughest week till now.This week I worked on plotScores function. plotScores function is a combination of 4 functions working together to make the plot. It was a little difficult to get started with it as the code was more complex than the rest of the functions. Once I understood the code of the function, I started with plotting the points. This is a special function as it used two kinds of symbol modes. In case of use.sym==FALSE the function plots a plot with same symbols but different colors, representing differnt spectra. On the other hand use.sym==TRUE plots a black and white plot with different spectra representing different symbols. After plotting the points I worked on plotting the ellipses. Initially I decided to to use the stat_ellipse geom, but later decied to use geom_path over it. Stat_ellipse basically calculates the coordinates of the ellipse and then plot them, however in our case we already had the coordinates precalculated, so using stat_ellipse seemed unnecessary. There are three options for plotting an ellipse in plotScores. First is classical way, second is robust way and the third way is plotting both of them together. I coded the ellipses one by one. After coding th plot my main focus was to use somehow use the leg.loc (legend location) used in base mode in ggplot2 mode. Sadly, ggplot2 does not provide a way of directly putting the legend in the desired position such as topright or bottomright etc . I manually coded the conditions in ggplot2 mode. Finally I worked on putting the ellipse information and method on the plot along with labelling of the points.

With this function ggplot2 mode is comming to an end with final testing and documentation remaining. By the end of next week I will hopefully start my work on plotly mode.

My commits during this period :

Commits of this week will be in the the pull request which will be merged next week.

Week 7

This week started with me completing the plotScores function in ChemoSpecUtils. The function was merged successfully. With this function the major part of ggplot2 mode is completed. While reviewing the code Bryan found a lot of functions where the code was getting repeated and could be simplified so I started my work on such functions.The main idea behind simplification was to create a function inside of the function, so instead of writing many lines of code a simple call to the function will do the job.I worked on plot2Loadings and many other other functions. After that I spent most of the time fixing small-small bugs in the functions. Legend layout was a little cramped so I had to work on that too. This week I learned about Rmarkdown too. I made some changes in the development version of the Rmakdown file.

Overall, there were a lot of minor tasks were worked on and completed throughout the week. Next week will mark the beginning of plotly mode!

My commits during this period :

Week 8

After giving the final touches in the ggplot2 mode, I started my work on the plotly mode. The very first thing was to learn how to combine the ggplot and plotly code together. plotly provides a function specifically for ggplot2 plot, ggplotly. The very first thing to do was to include the necessary imports for plotly mode. After adding the imports I worked on the very first plotSpectra function. At first I decided to use the seperate if conditions for plotly mode, however most of the code was getting repeated so Bryan suggested to put ggplot2 and plotly mode together, greatly reducing the code size. Since it was the first function in plotly mode I tried adding new stuff to it. After that I did some work on the surveySpectra and surveySpectra2 function. Overall, not much work was done in the first half of the week as a lot of the time was spent on reading the documentation.

With the end of this week, I move into the final and probably the most exciting phase of GSOC!

My commits during this period :

https://github.com/bryanhanson/ChemoSpec/pull/71/commits

Week 9

This week was a really long week. I started to work on the remaining functions of the ChemoSpec and ChemoSpecUtils. I first completed the surveySpectra and surveySpectra2 functions from the previous week. After that I worked on reviewAllSpectra, ploat2Loadings,sPlotSpectra, plotSpectraDist and finally pcaDiag. Adding plotly mode to all these functions was a similar process. I encountered a problem while adding plotly mode in plotLoadings function. The plotLoading function uses a patchwork package to put together all the plots in a grid. patchwork package works great with ggplot2 plots. However, it is yet to be implemented for plotly plots, so after a little research we decided to pause the work on plotLoadings function. After completing the plotly mode in ChemoSpec package, I decided to work on the ChemoSpecUtils functions. Bryan observed that some code was getting repeated in many functions, so my first task was to create a new function which could replace all the lengthy code into a single function call. I created .prepLegendCoords function to solve the issue. After that I worked on the previous listed issue in plotScore function regarding x and y axes labels. After that Bryan came up with an idea to further reduce the extra code by creating a specific legend function(.ggAddLegend) for ggplot2 mode. These both functoin together have removed a lot of repeated code and will be useful for further functions.

With this week over, we move on to the final week of the coding period!

My commits during this period :

https://github.com/bryanhanson/ChemoSpec/pull/72/commits
https://github.com/bryanhanson/ChemoSpec/pull/73/commits
https://github.com/bryanhanson/ChemoSpec/pull/74/commits
https://github.com/bryanhanson/ChemoSpecUtils/pull/9/commits

Week 10

This week marks the end of the coding period.Ten weeks long journey final comes to end. This week I first added the newly created .ggAddLegend function in .scorePlot and then in plotSpectra. After that I worked on adding plotly graphics mode in plotscores,aovPCAcores and plotScree function. I did stumbled in the middle when I tried to change the default graphics mode from base to ggplot2.After lot’s of searching and debugging I was able to reset it again to base graphics mode. After finishing all the plotting functions I worked on the small bugs that crept in due to me switching between the ChemoSpec and ChemoSpecUtils. After removing the bugs I updated the graphics option back to ggplot2 mode. With this I finally completed my work!

My commits during this period :

https://github.com/bryanhanson/ChemoSpec/pull/75/commits
https://github.com/bryanhanson/ChemoSpec/pull/76/commits
https://github.com/bryanhanson/ChemoSpecUtils/pull/10/commits
https://github.com/bryanhanson/ChemoSpecUtils/pull/11/commits
https://github.com/bryanhanson/ChemoSpecUtils/pull/12/commits
https://github.com/bryanhanson/ChemoSpecUtils/pull/13/commits
https://github.com/bryanhanson/ChemoSpecUtils/pull/14/commits

My Overall Experience

Participating in GSOC has been an amazing journey. I worked on different plotting functions, used differnt R packages and gained a lot of experience by working under my mentor. There were a lot times when I was lost and even behind the deadline but I was able to figure out the things and complete my work in the end. My mentor, Bryan, helped me throughout the journey and it would have been impossible to finish all this without him.

My whole coding journey can be summarised as:

Understand the previously written code.
Write the new code.
Test the code.
Try to come up with optimal solution to minimise the time and space complexity.
If some code is getting repeated, create a helper function to remove the excess code.
Finally write documentation of the code.
Repeat.

I hope that the new graphics modes raise the bar of ChemoSpec package and benefit the community who uses this package.

Special mention:

More about ChemoSpec can be found here.

Project Description

Mentors

Tasks completed

New imports included in ChemoSpec

Example plots with patchwork,ggrepel and ggplot2 packages

What’s done!

What’s next for the project?!

Weekly Blogs with Commits :

Community Bonding period

My commits during this period :

Week 1

My commits during this period :

Week 2

My commits during this period :

Week 3

My commits during this period :

Week 4

My commits during this period :

Week 5

My commits during this period :

Week 6

My commits during this period :

Week 7

My commits during this period :

Week 8

My commits during this period :

Week 9

My commits during this period :

Week 10

My commits during this period :

My Overall Experience

My whole coding journey can be summarised as:

Special mention:

Example plots with `patchwork`,`ggrepel` and `ggplot2` packages