Plotly bars with x-axis offset for timestamps - r

Question

I want to plot some timestamps with plotly bars, with 1 bar indicating a whole hour.
My problem is that the ticks are centered in the middle and I would like to shift them to the left end of the bars.
When the plot isn't zoomed in, it's not such a problem, but when zooming in, more tick-labels will appear and they would be wrong.
EDIT: I need the option barmode = 'overlay' as I also have other traces to plot, which are not included in this example.
The picture below illustrates my current and exptected layout and here's some data to make that plot. (Some option I tried without success are also included in the xaxis configuration but uncommented).
library(plotly)
library(lubridate)
df <- data.frame(
ts = seq(as.POSIXct("2019-03-20 00:00:00"), by = "hour", length.out = 24),
val = sample(1:100, 24)
)
plot_ly() %>%
add_bars(data = df, x = ~ts, y = ~val) %>%
layout(dragmode = "select", autosize = TRUE, selectdirection = "h",
barmode = 'overlay',
bargap = 0.05,
xaxis = list(ticks = "outside",
type = "date",
# tickson="boundaries",
# offset=1800,
tickmode = "auto",
title = ""
)) %>%
config(scrollZoom = TRUE)

Answer 1

Would the following meet your needs?
One of the things that I take advantage of sometimes with plotly is that you can show different values in text that are independent of your the x and y values used to plot the data.
In this case, we can create a column with an offset time value, ts_x and plot the x values a half hour past the time for each row -- If you have a column for every hour, this effectively left-aligns the bars.
library(plotly)
df <- data.frame(
ts = seq(as.POSIXct("2019-03-20 00:00:00"), by = "hour", length.out = 24),
val = sample(1:100, 24)
)
## Create a dummy column with x offset values
df$ts_x <- df$ts + 1800
plot_ly() %>%
## Plot based on the dummy column
add_bars(data = df, x = ~ts_x, y = ~val,
## Cover up our tracks by not showing true x value on hoverinfo
hoverinfo = "text",
## Give text that includes the un-altered time values
text = ~paste0("Time: ",format(ts, format = "%B %d, %Y %H:%M"),
"<br>Value: ",val)) %>%
layout(dragmode = "select", autosize = TRUE, selectdirection = "h",
barmode = 'overlay',
bargap = 0.05,
xaxis = list(ticks = "outside",
type = "date",
tickmode = "auto",
title = ""
)) %>%
config(scrollZoom = TRUE)

Answer 2

By default, bars are centered, I didn't find how to change this.
One alternative is to add a second bar, because when there are 2 bars for each x-axis unity, one bar is at the left of the axis tick, and the second at the right (what you are trying to obtain with one bar).
Why not creating a second invisible bar ? :)
df <- data.frame(
ts = seq(as.POSIXct("2019-03-20 00:00:00"), by = "hour", length.out = 24),
val = sample(1:100, 24),
val0 = 0
)
plot_ly(df, type = 'bar') %>%
add_trace(x = ~ts, y = ~val0) %>%
add_trace(x = ~ts, y = ~val) %>%
layout(
showlegend = FALSE
) %>%
config(scrollZoom = TRUE)
This will create a legend (as there are 2 kind of bars, ones for val and ones for val0), so I removed it.

Answer 3

Are you sure you are not over engineering? Subtracting 30 minutes gives me a nice graph when zooming in.
I'm not suggest you actually edit the data, even if it's what I'm doing in the code. A small function in the call to add bars could solve it? If you overlay other data it could make a mess but I just wanted to suggest it.
library(plotly)
library(lubridate)
df2 <- data.frame(
ts = seq(as.POSIXct("2019-03-20 00:00:00"), by = "hour", length.out = 24) - minutes(30),
val = sample(1:100, 24)
)
plot_ly() %>%
add_bars(data = df2, x = ~ts, y = ~val) %>%
layout(dragmode = "select", autosize = TRUE, selectdirection = "h",
barmode = 'overlay',
bargap = 0.05,
xaxis = list(ticks = "outside",
type = "date",
# tickson="boundaries",
# offset=1800,
tickmode = "auto",
title = ""
)) %>%
config(scrollZoom = TRUE)