Should remove breaks for OHLC, Candlestick

[mckelvin]

The code to reproduce is based on simple-candlestick-chart-example . The difference is:

11th, 12th are not existed in the dataset (simply because they're weekdays/non-trading days), but candlesticks in the chart should be continued, i.e.: the breaks of 11th, 12th should be removed like what highcharts do.

This behaviour should be the default one for OHLC and Candlestick charts, otherwise it will become terrible for real world dataset.

---

https://codepen.io/mckelvin/pen/ygwMQz is my workaround by using tickvals and ticktext. But there's another problem: nticks doesn't work when tickmode=array as mentioned in docs, there will be too many tick labels if there're too many ticks.

[mckelvin]

@etpinard Thanks for your reply. Sorry but you might misunderstand me. I don't mean to argue for the feature as it's very common in OHLC/Candlestick charts and the link is just provided as a demo which contains the feature I expected. Both plotly and highcharts are great softwares but neither of them is a copycat.

[jackparmer]

Current proposal is to either add a new axis type:
type: 'date-no-we'

Or a flag for the axis object:
show_weekend: false

cc @jackwluo @cpsievert @santosjorge

Anyone out there interested in sponsoring this one? 🥂
https://plot.ly/products/consulting-and-oem/
(Or even better, submit a PR)

[jackluo]

Implement is as a 'financial' flag for axis, since different markets have different holidays

[jackparmer]

@jackwluo - That's an interesting point. Can you think of any online reference for which markets respect which holidays?

It looks like mpl and ggplot user have also come across this and have their own workarounds:

ggplot
mpl

[jackluo]

Seems like NYTimes has a comprehensive database;

http://markets.on.nytimes.com/research/markets/holidays/holidays.asp?display=market&exchange=LSE

Or individual exchange links below

https://www.nyse.com/markets/hours-calendars
https://www.tsx.com/trading/calendars-and-trading-hours/calendar-and-events
http://www.jpx.co.jp/english/corporate/calendar/

[mckelvin]

This issue not only happens on dates, but also on other resolutions like 1min, 5min, 1h since the trading session in a day might be 08:00~16:00 (not full 24h). So I think the library should not care about the detailed holidays.

SEE ALSO: https://www.tradingview.com

[santosjorge]

In my opinion - I think it is an overkill to expect Plotly having to retrieve holidays/trading sessions for all instruments. You will also have to handle scenarios where there is multiple calendars (ie a GBP swap).

Most data providers already take into account non-trading days/market sessions into account. That means the data has already been cleaned/adjusted.

The only thing required would be a layout parameter to create an axis that will omit the dates for which no values exist in the timeseries.

[Feng-Zihao]

It's not necessary to retrive all holidays. Every stock market, every country ,may have their own holidays.

Just treat the datetime x_Axis as 'string' instead of 'Date'.

Any update on this? I think a general solution would be an option to ignore missing dates. Other solutions require avoiding use of dates for the x-axis which is a big handicap for features eg use rangeslider/rangeselector won't work for array.

[etpinard]

This won't be priority for us until we get a sponsor for it:

https://plot.ly/products/consulting-and-oem/

[mckelvin]

@etpinard Make sense but I don't agree with the priority. IMHO, the missing feature is the largest barrier ahead of the financial users since the OHLC chart is almost the basic chart of all other high level indicator charts. If they can't even try plotly to draw the basic chart, you can neither expect them to use it and pay for consulting.

More and more financial/trading companies start to use python stack for data analysis (at least I know Optiver is using). Plotly works so well on Jupyter. They are generous to pay for what could save time and bring profit. I'd happy to see good match between plotly and trading companies. That's really a good reason to increase the priority. Please reconsider it.

I'd agree. If the basic price chart is broken, it's just not a viable solution - it's a real impediment to people even trying.