-Wlto-type-mismatch warnings in CRAN's LTO check.This version brings many changes to the R package. The entire code generating backend have been rewritten so it closely follows the upstream naming of parameters and it uses X-macros so it also installs way fastert than before - but it also means that there is alot of breaking changes. The update is a big leap towards a stable release.
talib::lookback(
FUN = talib::SMA,
timePeriod = 10,
x = talib::BTC
)
#> [1] 9
The function returns the required lookback for calculating the indicator. A lookback of 0 means the indicator is computable from the first observation (talib::balance_of_power(), for example).
Its use-case is customized control-flows for downstream wrappers and/or packages that declares dependency on {talib}.
lookback attribute: the total number of leading rows that are not computed. The attribute is cumulative, so chaining indicators sums each stage's lookback, and it is used internally to chain indicators deterministically at full input length:tail(
output <- talib::BBANDS(
talib::SMA(
talib::RSI(
talib::GOOGL
)
)
)
)
#> UpperBand MiddleBand LowerBand
#> 2021-12-22 57.50571 54.51537 51.52503
#> 2021-12-23 57.26401 54.23193 51.19985
#> 2021-12-27 57.02005 53.97461 50.92918
#> 2021-12-28 56.75847 53.71577 50.67306
#> 2021-12-29 56.55065 53.47299 50.39534
#> 2021-12-30 56.40837 53.25097 50.09356
stopifnot(
nrow(talib::GOOGL) == nrow(output)
)
attr(
output,
"lookback"
)
#> [1] 62
The chain carries RSI (14), SMA (29) and BBANDS (19)—attr(output, "lookback") is their sum, the leading-NA head of the final series. The per-indicator lookback remains available via talib::lookback().
The source code have been re-written so it generates the underlying TA-Lib wrappers using preprocessors and X-Macros, which compiles much faster than before.
<xts>-methods:—All indicators now supports <xts>-objects. These methods are considered the primary entry point for all indicators, and can be considered stable from v1.0.0, where changes—if any—will be implemented gradually after a deprecation period. The method uses the same signature as before, see the example below:
library(xts)
tail(
x <- talib::bollinger_bands(
talib::GOOGL
)
)
#> UpperBand MiddleBand LowerBand
#> 2021-12-22 149.1432 144.4920 139.8408
#> 2021-12-23 149.2513 144.5318 139.8124
#> 2021-12-27 149.6263 144.8180 140.0097
#> 2021-12-28 149.7444 144.8758 140.0072
#> 2021-12-29 149.8398 145.1137 140.3876
#> 2021-12-30 149.7308 145.3711 141.0115
class(x)
#> [1] "xts" "zoo"
talib::variable_moving_average_period(
x = 1:10,
periods = c(1, 1, 1, 2, 2, 2, 3, 4, 4, 4),
minimumPeriod = 2,
maximumPeriod = 4
)
#> [1] NA NA NA 3.5 4.5 5.5 6.0 6.5 7.5 8.5
#> attr(,"lookback")
talib::average_deviation(
x = 1:10,
periods = c(1, 1, 1, 2, 2, 2, 3, 4, 4, 4),
timePeriod = 5
)
#> [1] NA NA NA NA 1.2 1.2 1.2 1.2 1.2 1.2
#> attr(,"lookback")
#> [1] 4
tail(
talib::CMF(
x = talib::GOOGL,
timePeriod = 20L,
cols = ~ high + low + close + volume
)
)
#> CMF
#> 2021-12-22 0.08511277
#> 2021-12-23 0.03979116
#> 2021-12-27 0.08378773
#> 2021-12-28 0.05341346
#> 2021-12-29 0.12197905
#> 2021-12-30 0.16362414
tail(
talib::CMOU(
x = talib::GOOGL,
timePeriod = 20L,
cols = ~close
)
)
#> CMOU
#> 2021-12-22 1.650705
#> 2021-12-23 2.068290
#> 2021-12-27 16.093739
#> 2021-12-28 3.458882
#> 2021-12-29 15.946284
#> 2021-12-30 17.488048
tail(
talib::HMA(
talib::GOOGL,
timePeriod = 20L,
cols = ~open
)
)
#> HMA
#> 2021-12-22 143.4272
#> 2021-12-23 143.1059
#> 2021-12-27 143.3396
#> 2021-12-28 144.0303
#> 2021-12-29 144.8070
#> 2021-12-30 145.5167
tail(
talib::NVI(
talib::GOOGL,
cols = ~ close + volume
)
)
#> NVI
#> 2021-12-22 3087.244
#> 2021-12-23 3087.244
#> 2021-12-27 3108.047
#> 2021-12-28 3108.047
#> 2021-12-29 3107.369
#> 2021-12-30 3097.739
tail(
talib::PVI(
talib::GOOGL,
cols = ~ close + volume
)
)
#> PVI
#> 2021-12-22 899.3401
#> 2021-12-23 902.4205
#> 2021-12-27 902.4205
#> 2021-12-28 894.9800
#> 2021-12-29 894.9800
#> 2021-12-30 894.9800
tail(
talib::VWMA(
x = talib::GOOGL,
timePeriod = 20L,
cols = ~ close + volume
)
)
#> VWMA
#> 2021-12-22 144.1797
#> 2021-12-23 144.2380
#> 2021-12-27 144.4321
#> 2021-12-28 144.4406
#> 2021-12-29 144.6905
#> 2021-12-30 144.9503
camelCase alias to introduce a form of consistency across R's finance ecosystem and oldschool coding schemes.
The indicators below produces the same output:talib::bollinger_bands()
talib::BBANDS()
talib::bollingerBands()
Each UPPERCASE and camelCase function is an alias of its underlying snake_case function, so the functions behaves the same.
indicator(
x, ## unchanged
cols, ## unchanged
## additional/optional TA-Lib parameters
## are now camelCase mined upstream
timePeriod, ## was 'n' before
fooBar, ## was 'foo_bar' or 'foobar' before
fooBaz, ## was 'foo_baz' or 'foobaz' before
na.bridge = FALSE ## unchanged
)
This has the benefit of being transparent when comparing or reading the source code.
bollinger_bands() below:talib::bollinger_bands(
talib::BTC,
timePeriod = 20,
maType = talib::EMA()
)
Prior to this update, the correct call was:
talib::bollinger_bands(
talib::BTC,
ma = talib::EMA(n = 20)
)
While the above function call is aestethically pleasing, it did introduce some ambigiuites in other calls. See, for example, APO() (v0.9.2) below:
absolute_price_oscillator(
x,
cols,
fast = 12,
slow = 26,
ma = SMA(n = 9),
na.bridge = FALSE,
...
)
In this specific case the function has three different n - the underlying function were discarding n = 9 while keeping the MAType.
The new call is given as:
absolute_price_oscillator(
x,
cols,
fastPeriod = 12,
slowPeriod = 26,
maType = 0,
na.bridge = FALSE,
...
)
In this call the role of each argument is should be clearer than before.
x + timePeriod signature, and breaks the numeric-method.
One solution is to do the following:VWMA.numeric(x, cols, timePeriod, na.bridge = FALSE, volume, ...)
To avoid breaking the S3 signature. However, given the importance of the volume-argument and the fact that cols is rarely used, the signature has been changed to accommodate future additions of MAs which take additional series. The new signature is as follows:
foo(x, [series], timePeriod, [optional], cols, na.bridge = FALSE, ...)
Additional series (the volume) lead directly after x, while additional optional parameters keep their place after timePeriod. In the case of VWMA this becomes:
VWMA(
x,
volume,
timePeriod = 30,
cols,
na.bridge = FALSE,
...
)
And for the remaining MAs:
foo(
x,
timePeriod = 30,
cols,
na.bridge = FALSE,
...
)
MAs with additional optional parameters retain them after timePeriod, e.g. MAMA(x, timePeriod, fastLimit, slowLimit, cols, na.bridge, ...) and T3(x, timePeriod, volumeFactor, cols, na.bridge, ...).
The volume-argument is not required when x carries a volume column: VWMA(x) selects it via the default formula ~close + volume as before. An explicitly passed vector takes precedence, in which case only the close-column is required:
## volume from the 'volume'-column of x
VWMA(x)
## explicitly passed volume; only 'close' required
VWMA(x, volume = my_volume)
## fully positional on vectors
VWMA(price, volume, 20)
NOTE: passing cols positionally as the second argument no longer works for the MAs—it has to be passed by name, ie. SMA(x, cols = ~open).
CCI: Incorrect charting—The indicator were incorrectly classified as a main chart indicator—
One-dimensional indicators: incorrect return <class>—Indicators that returns a one-dimensional indicator (MA, RSI, etc.) were returning a <matrix> or <data.frame> instead of <numeric>.
Merged indicators: overlapping last-values—On the plotly backend, merging indicators onto one panel stacked every last-value label on the panel's top-right corner. The labels are now collapsed into a single evenly spaced label, mirroring the merged subtitle of the ggplot2 backend.
Merged indicators: indistinguishable last-values—Merged last-value labels used the bare output column name, so merging the same indicator with different parameters displayed identical labels. The labels now carry the full indicator specification, including its parameters, as the ggplot2 backend already did.
The configure for UNIX have been improved and follows 'Writing R Extensions' more closely so its more robust across different operating systems.
The configure now probes for user-installed libraries uses pkgconf and should locate libraries and headers installed in non-default PATHs.
The MAMA-function now has two arguments: fast and slow, which controls
controls the upper and lower limit of the adaptive smoothing factor (alpha) used
in the MESA algorithm
The T3-function now has the argument vfactor controls the smoothing weight of the T3-curve.
stochastic_relative_strength_index-function were recursively calculating the indicator.chart() and indicator(), with both
plotly (interactive) and ggplot2 (static) backends.BTC, ATOM, NVDA, SPY.