NB: This vignette is work-in-progress and not yet complete.
Short intro on hashes
TBD
Difference between digest()
and sha1()
R FAQ 7.31 illustrates potential problems with floating point arithmetic. Mathematically the equality \(x = \sqrt{x}^2\) should hold. But the precision of floating points numbers is finite. Hence some rounding is done, leading to numbers which are no longer identical.
An illustration:
# FAQ 7.31
<- 2
a0 <- sqrt(a0)
b <- b ^ 2
a1 identical(a0, a1)
## [1] FALSE
- a1
a0 ## [1] -4.440892e-16
<- c(a0, a1)
a # hexadecimal representation
sprintf("%a", a)
## [1] "0x1p+1" "0x1.0000000000001p+1"
Although the difference is small, any difference will result in different hash when using the digest()
function. However, the sha1()
function tackles this problem by using the hexadecimal representation of the numbers and truncates that representation to a certain number of digits prior to calculating the hash function.
library(digest)
# different hashes with digest
sapply(a, digest, algo = "sha1")
## [1] "315a5aa84aa6cfa4f3fb4b652a596770be0365e8"
## [2] "5e3999bf79c230f7430e97d7f30070a7eff5ec92"
# same hash with sha1 with default digits (14)
sapply(a, sha1)
## [1] "8a938d8f5fb9b8ccb6893aa1068babcc517f32d4"
## [2] "8a938d8f5fb9b8ccb6893aa1068babcc517f32d4"
# larger digits can lead to different hashes
sapply(a, sha1, digits = 15)
## [1] "98eb1dc9fada00b945d3ef01c77049ee5a4b7b9c"
## [2] "5a173e2445df1134908037f69ac005fbd8afee74"
# decreasing the number of digits gives a stronger truncation
# the hash will change when then truncation gives a different result
# case where truncating gives same hexadecimal value
sapply(a, sha1, digits = 13)
## [1] "43b3b465c975af322c85473190a9214b79b79bf6"
## [2] "43b3b465c975af322c85473190a9214b79b79bf6"
sapply(a, sha1, digits = 10)
## [1] "6b537a9693c750ed535ca90527152f06e358aa3a"
## [2] "6b537a9693c750ed535ca90527152f06e358aa3a"
# case where truncating gives different hexadecimal value
c(sha1(pi), sha1(pi, digits = 13), sha1(pi, digits = 10))
## [1] "169388cf1ce60dc4e9904a22bc934c5db33d975b"
## [2] "20dc38536b6689d5f2d053f30efb09c44af78378"
## [3] "3a727417bd1807af2f0148cf3de69abff32c23ec"
The result of floating point arithematic on 32-bit and 64-bit can be slightly different. E.g. print(pi ^ 11, 22)
returns 294204.01797389047
on 32-bit and 294204.01797389053
on 64-bit. Note that only the last 2 digits are different.
command | 32-bit | 64-bit |
---|---|---|
print(pi ^ 11, 22) |
294204.01797389047 |
294204.01797389053 |
sprintf("%a", pi ^ 11) |
"0x1.1f4f01267bf5fp+18" |
"0x1.1f4f01267bf6p+18" |
digest(pi ^ 11, algo = "sha1") |
"c5efc7f167df1bb402b27cf9b405d7cebfba339a" |
"b61f6fea5e2a7952692cefe8bba86a00af3de713" |
sha1(pi ^ 11, digits = 14) |
"5c7740500b8f78ec2354ea6af58ea69634d9b7b1" |
"4f3e296b9922a7ddece2183b1478d0685609a359" |
sha1(pi ^ 11, digits = 13) |
"372289f87396b0877ccb4790cf40bcb5e658cad7" |
"372289f87396b0877ccb4790cf40bcb5e658cad7" |
sha1(pi ^ 11, digits = 10) |
"c05965af43f9566bfb5622f335817f674abfc9e4" |
"c05965af43f9566bfb5622f335817f674abfc9e4" |
Choosing digest()
or sha1()
TBD
Creating a sha1 method for other classes
How to
- Identify the relevant components for the hash.
- Determine the class of each relevant component and check if they are handled by
sha1()
.- Write a method for each component class not yet handled by
sha1
.
- Write a method for each component class not yet handled by
- Extract the relevant components.
- Combine the relevant components into a list. Not required in case of a single component.
- Apply
sha1()
on the (list of) relevant component(s). - Turn this into a function with name sha1._classname_.
- sha1._classname_ needs exactly the same arguments as
sha1()
- Choose sensible defaults for the arguments
zapsmall = 7
is recommended.digits = 14
is recommended in case all numerics are data.digits = 4
is recommended in case some numerics stem from floating point arithmetic.
summary.lm
Let’s illustrate this using the summary of a simple linear regression. Suppose that we want a hash that takes into account the coefficients, their standard error and sigma.
# taken from the help file of lm.influence
<- lm(sr ~ pop15 + pop75 + dpi + ddpi, data = LifeCycleSavings)
lm_SR <- summary(lm_SR)
lm_sum class(lm_sum)
## [1] "summary.lm"
# str() gives the structure of the lm object
str(lm_sum)
## List of 11
## $ call : language lm(formula = sr ~ pop15 + pop75 + dpi + ddpi, data = LifeCycleSavings)
## $ terms :Classes 'terms', 'formula' language sr ~ pop15 + pop75 + dpi + ddpi
## .. ..- attr(*, "variables")= language list(sr, pop15, pop75, dpi, ddpi)
## .. ..- attr(*, "factors")= int [1:5, 1:4] 0 1 0 0 0 0 0 1 0 0 ...
## .. .. ..- attr(*, "dimnames")=List of 2
## .. .. .. ..$ : chr [1:5] "sr" "pop15" "pop75" "dpi" ...
## .. .. .. ..$ : chr [1:4] "pop15" "pop75" "dpi" "ddpi"
## .. ..- attr(*, "term.labels")= chr [1:4] "pop15" "pop75" "dpi" "ddpi"
## .. ..- attr(*, "order")= int [1:4] 1 1 1 1
## .. ..- attr(*, "intercept")= int 1
## .. ..- attr(*, "response")= int 1
## .. ..- attr(*, ".Environment")=<environment: R_GlobalEnv>
## .. ..- attr(*, "predvars")= language list(sr, pop15, pop75, dpi, ddpi)
## .. ..- attr(*, "dataClasses")= Named chr [1:5] "numeric" "numeric" "numeric" "numeric" ...
## .. .. ..- attr(*, "names")= chr [1:5] "sr" "pop15" "pop75" "dpi" ...
## $ residuals : Named num [1:50] 0.864 0.616 2.219 -0.698 3.553 ...
## ..- attr(*, "names")= chr [1:50] "Australia" "Austria" "Belgium" "Bolivia" ...
## $ coefficients : num [1:5, 1:4] 28.566087 -0.461193 -1.691498 -0.000337 0.409695 ...
## ..- attr(*, "dimnames")=List of 2
## .. ..$ : chr [1:5] "(Intercept)" "pop15" "pop75" "dpi" ...
## .. ..$ : chr [1:4] "Estimate" "Std. Error" "t value" "Pr(>|t|)"
## $ aliased : Named logi [1:5] FALSE FALSE FALSE FALSE FALSE
## ..- attr(*, "names")= chr [1:5] "(Intercept)" "pop15" "pop75" "dpi" ...
## $ sigma : num 3.8
## $ df : int [1:3] 5 45 5
## $ r.squared : num 0.338
## $ adj.r.squared: num 0.28
## $ fstatistic : Named num [1:3] 5.76 4 45
## ..- attr(*, "names")= chr [1:3] "value" "numdf" "dendf"
## $ cov.unscaled : num [1:5, 1:5] 3.74 -7.24e-02 -4.46e-01 -7.86e-05 -1.88e-02 ...
## ..- attr(*, "dimnames")=List of 2
## .. ..$ : chr [1:5] "(Intercept)" "pop15" "pop75" "dpi" ...
## .. ..$ : chr [1:5] "(Intercept)" "pop15" "pop75" "dpi" ...
## - attr(*, "class")= chr "summary.lm"
# extract the coefficients and their standard error
<- coef(lm_sum)[, c("Estimate", "Std. Error")]
coef_sum # extract sigma
<- lm_sum$sigma
sigma # check the class of each component
class(coef_sum)
## [1] "matrix" "array"
class(sigma)
## [1] "numeric"
# sha1() has methods for both matrix and numeric
# because the values originate from floating point arithmetic it is better to use a low number of digits
sha1(coef_sum, digits = 4)
## [1] "3f0b0c552f94d753fcc8deb4d3e9fc11a83197af"
sha1(sigma, digits = 4)
## [1] "cbc83d1791ef1eeadd824ea9a038891b5889056b"
# we want a single hash
# combining the components in a list is a solution that works
sha1(list(coef_sum, sigma), digits = 4)
## [1] "476d27265365cd41662eedf059b335d38a221cc2"
# now turn everything into an S3 method
# - a function with name "sha1.classname"
# - must have the same arguments as sha1()
<- function(x, digits = 4, zapsmall = 7){
sha1.summary.lm <- coef(x)[, c("Estimate", "Std. Error")]
coef_sum <- x$sigma
sigma <- list(coef_sum, sigma)
combined sha1(combined, digits = digits, zapsmall = zapsmall)
}sha1(lm_sum)
## [1] "476d27265365cd41662eedf059b335d38a221cc2"
# try an altered dataset
<- LifeCycleSavings[rownames(LifeCycleSavings) != "Zambia", ]
LCS2 <- lm(sr ~ pop15 + pop75 + dpi + ddpi, data = LCS2)
lm_SR2 sha1(summary(lm_SR2))
## [1] "90beb028833bf0542997fde7c3f19e5b9fdfeef4"
lm
Let’s illustrate this using the summary of a simple linear regression. Suppose that we want a hash that takes into account the coefficients, their standard error and sigma.
class(lm_SR)
## [1] "lm"
# str() gives the structure of the lm object
str(lm_SR)
## List of 12
## $ coefficients : Named num [1:5] 28.566087 -0.461193 -1.691498 -0.000337 0.409695
## ..- attr(*, "names")= chr [1:5] "(Intercept)" "pop15" "pop75" "dpi" ...
## $ residuals : Named num [1:50] 0.864 0.616 2.219 -0.698 3.553 ...
## ..- attr(*, "names")= chr [1:50] "Australia" "Austria" "Belgium" "Bolivia" ...
## $ effects : Named num [1:50] -68.38 -14.29 7.3 -3.52 -7.94 ...
## ..- attr(*, "names")= chr [1:50] "(Intercept)" "pop15" "pop75" "dpi" ...
## $ rank : int 5
## $ fitted.values: Named num [1:50] 10.57 11.45 10.95 6.45 9.33 ...
## ..- attr(*, "names")= chr [1:50] "Australia" "Austria" "Belgium" "Bolivia" ...
## $ assign : int [1:5] 0 1 2 3 4
## $ qr :List of 5
## ..$ qr : num [1:50, 1:5] -7.071 0.141 0.141 0.141 0.141 ...
## .. ..- attr(*, "dimnames")=List of 2
## .. .. ..$ : chr [1:50] "Australia" "Austria" "Belgium" "Bolivia" ...
## .. .. ..$ : chr [1:5] "(Intercept)" "pop15" "pop75" "dpi" ...
## .. ..- attr(*, "assign")= int [1:5] 0 1 2 3 4
## ..$ qraux: num [1:5] 1.14 1.17 1.16 1.15 1.05
## ..$ pivot: int [1:5] 1 2 3 4 5
## ..$ tol : num 1e-07
## ..$ rank : int 5
## ..- attr(*, "class")= chr "qr"
## $ df.residual : int 45
## $ xlevels : Named list()
## $ call : language lm(formula = sr ~ pop15 + pop75 + dpi + ddpi, data = LifeCycleSavings)
## $ terms :Classes 'terms', 'formula' language sr ~ pop15 + pop75 + dpi + ddpi
## .. ..- attr(*, "variables")= language list(sr, pop15, pop75, dpi, ddpi)
## .. ..- attr(*, "factors")= int [1:5, 1:4] 0 1 0 0 0 0 0 1 0 0 ...
## .. .. ..- attr(*, "dimnames")=List of 2
## .. .. .. ..$ : chr [1:5] "sr" "pop15" "pop75" "dpi" ...
## .. .. .. ..$ : chr [1:4] "pop15" "pop75" "dpi" "ddpi"
## .. ..- attr(*, "term.labels")= chr [1:4] "pop15" "pop75" "dpi" "ddpi"
## .. ..- attr(*, "order")= int [1:4] 1 1 1 1
## .. ..- attr(*, "intercept")= int 1
## .. ..- attr(*, "response")= int 1
## .. ..- attr(*, ".Environment")=<environment: R_GlobalEnv>
## .. ..- attr(*, "predvars")= language list(sr, pop15, pop75, dpi, ddpi)
## .. ..- attr(*, "dataClasses")= Named chr [1:5] "numeric" "numeric" "numeric" "numeric" ...
## .. .. ..- attr(*, "names")= chr [1:5] "sr" "pop15" "pop75" "dpi" ...
## $ model :'data.frame': 50 obs. of 5 variables:
## ..$ sr : num [1:50] 11.43 12.07 13.17 5.75 12.88 ...
## ..$ pop15: num [1:50] 29.4 23.3 23.8 41.9 42.2 ...
## ..$ pop75: num [1:50] 2.87 4.41 4.43 1.67 0.83 2.85 1.34 0.67 1.06 1.14 ...
## ..$ dpi : num [1:50] 2330 1508 2108 189 728 ...
## ..$ ddpi : num [1:50] 2.87 3.93 3.82 0.22 4.56 2.43 2.67 6.51 3.08 2.8 ...
## ..- attr(*, "terms")=Classes 'terms', 'formula' language sr ~ pop15 + pop75 + dpi + ddpi
## .. .. ..- attr(*, "variables")= language list(sr, pop15, pop75, dpi, ddpi)
## .. .. ..- attr(*, "factors")= int [1:5, 1:4] 0 1 0 0 0 0 0 1 0 0 ...
## .. .. .. ..- attr(*, "dimnames")=List of 2
## .. .. .. .. ..$ : chr [1:5] "sr" "pop15" "pop75" "dpi" ...
## .. .. .. .. ..$ : chr [1:4] "pop15" "pop75" "dpi" "ddpi"
## .. .. ..- attr(*, "term.labels")= chr [1:4] "pop15" "pop75" "dpi" "ddpi"
## .. .. ..- attr(*, "order")= int [1:4] 1 1 1 1
## .. .. ..- attr(*, "intercept")= int 1
## .. .. ..- attr(*, "response")= int 1
## .. .. ..- attr(*, ".Environment")=<environment: R_GlobalEnv>
## .. .. ..- attr(*, "predvars")= language list(sr, pop15, pop75, dpi, ddpi)
## .. .. ..- attr(*, "dataClasses")= Named chr [1:5] "numeric" "numeric" "numeric" "numeric" ...
## .. .. .. ..- attr(*, "names")= chr [1:5] "sr" "pop15" "pop75" "dpi" ...
## - attr(*, "class")= chr "lm"
# extract the model and the terms
<- lm_SR$model
lm_model <- lm_SR$terms
lm_terms # check their class
class(lm_model) # handled by sha1()
## [1] "data.frame"
class(lm_terms) # not handled by sha1()
## [1] "terms" "formula"
# define a method for formula
<- function(x, digits = 14, zapsmall = 7, ..., algo = "sha1"){
sha1.formula sha1(as.character(x), digits = digits, zapsmall = zapsmall, algo = algo)
}sha1(lm_terms)
## [1] "2737d209720aa7d1c0555050ad06ebe89f3850cd"
sha1(lm_model)
## [1] "27b7dd9e3e09b9577da6947b8473b63a1d0b6eb4"
# define a method for lm
<- function(x, digits = 14, zapsmall = 7, ..., algo = "sha1"){
sha1.lm <- x$model
lm_model <- x$terms
lm_terms <- list(lm_model, lm_terms)
combined sha1(combined, digits = digits, zapsmall = zapsmall, ..., algo = algo)
}sha1(lm_SR)
## [1] "7eda2a9d58e458c8e782e40ce140d62b836b2a2f"
sha1(lm_SR2)
## [1] "4d3abdb1f17bd12fdf9d9b91a2ad04c07824fe4a"
Using hashes to track changes in analysis
Use case
automated analysis
update frequency of the data might be lower than the frequency of automated analysis
similar analyses on many datasets (e.g. many species in ecology)
analyses that require a lot of computing time
- not rerunning an analysis because nothing has changed saves enough resources to compensate the overhead of tracking changes
Bundle all relevant information on an analysis in a class
- data
- method
- formula
- other metadata
- resulting model
calculate
sha1()
file fingerprint ~
sha1()
on the stable partsstatus fingerprint ~
sha1()
on the parts that result for the model
- Prepare analysis objects
- Store each analysis object in a rda file which uses the file fingerprint as filename
- File will already exist when no change in analysis
- Don’t overwrite existing files
- Loop over all rda files
- Do nothing if the analysis was run
- Otherwise run the analysis and update the status and status fingerprint