How to get the grouping right in R with Plotly - r

Question

I have some problem to group my data in Plotly under R. To start with I was using local data from a csv file, reading them with:
geogrid_data <- read.delim('geogrid.csv', row.names = NULL, stringsAsFactors = TRUE)
and the plotting went well, using the following:
library(plotly)
library(RColorBrewer)
x <- list(
title = 'Date'
)
p <- plotly::plot_ly(geogrid_data,
type = 'scatter',
x = ~ts_now,
y = ~absolute_v_sum,
text = paste('Table: ', geogrid_data$table_name,
'<br>Absolute_v_Sum: ', geogrid_data$absolute_v_sum),
hoverinfo = 'text',
mode = 'lines',
color = list(
color = colorRampPalette(RColorBrewer::brewer.pal(11,'Spectral'))(
length(unique(geogrid_data$table_name))
)
),
transforms = list(
list(
type = 'groupby',
groups = ~table_name
)
)
) %>% layout(showlegend = TRUE, xaxis = x)
Here the output:
Then I was going to alter the data source to an Oracle database table, reading the data as follows, using the ROracle package:
# retrieve data into resultSet object
rs <- dbSendQuery(con, "SELECT * FROM GEOGRID_STATS")
# fetch records from the resultSet into a data.frame
geogrid_data <- fetch(rs)
# free resources occupied by resultSet
dbClearResult(rs)
dbUnloadDriver(drv)
# remove duplicates from dataframe (based on TABLE_NAME, TS_BEFORE, TS_NOW, NOW_SUM)
geogrid_data <- geogrid_data %>% distinct(TABLE_NAME, TS_BEFORE, TS_NOW, NOW_SUM, .keep_all = TRUE)
# alter date columns in place
geogrid_data$TS_BEFORE <- as.Date(geogrid_data$TS_BEFORE, format='%d-%m-%Y')
geogrid_data$TS_NOW <- as.Date(geogrid_data$TS_NOW, format='%d-%m-%Y')
and adjusting the plotting to:
p <- plotly::plot_ly(
type = 'scatter',
x = geogrid_data$TS_NOW,
y = geogrid_data$ABSOLUTE_V_SUM,
text = paste('Table: ', geogrid_data$TABLE_NAME,
'<br>Absolute_v_Sum: ', geogrid_data$ABSOLUTE_V_SUM,
'<br>Date: ', geogrid_data$TS_NOW),
hoverinfo = 'text',
mode = 'lines',
color = list(
color = colorRampPalette(RColorBrewer::brewer.pal(11,'Spectral'))(
length(unique(geogrid_data$TABLE_NAME))
)
),
transforms = list(
list(
type = 'groupby',
groups = geogrid_data$TABLE_NAME
)
)
) %>% layout(showlegend = TRUE, xaxis = x)
Unfortunately, this is leading to some problem with the grouping as it seems.:
As you can see from the label text when hovering over the data point, the point represents data from NY_SKOV_PLANTEB_MW_POLY while the legend is set to show data from NY_BYGN_MW_POLY. Looking at other data points in this graph I found a wild mix of points of all sorts in this graph, some of them representing data of NY_BYGN_MW_POLY, most of them not.
Also the plotting with respect to the time line does not work any more, e.g. data are plotted with start on Dec. 11 - Dec. 10 - Dec. 10 - Dec. 12 - Dec. 20 - Dec. 17 - Dec. 16 - Dec. 15.
Where do I go wrong in handling the data, and what do I have to do to get it right?

Answer 1

Of course, one should look at the data... thanks Marco, after your question I did look at my data.
There are some points where I simply assumed things.
The reason why all data plotted fine with data from the csv file is simple. All information manually compiled in the csv file came from information in emails that have been ordered by date. Hence, I compiled the data in the csv file ordered by date and Plotly does not have any problems grouping the data by table_name.
After looking at my data I tidied up, keeping only the data I need to show in the plot and used dplyr to sort the data by time.
geogrid_data <- dplyr::arrange(geogrid_data, TS_NOW)
It is only by time and not by time and table name because the sorting by table name is done anyway by Plotly and the groupby statement