Guest *Charles Berry* @ 2012-08-18 04:08:00 originally posted:

Can the caching of chunks in child documents be made to depend on chunks in the parent document?

I have used dependson="mainchunk" in a child document, where 'mainchunk' appears in the parent. No error is reported, but the cache does not get rebuilt when the code in 'mainchunk' is altered.

Guest *Hyao* @ 2012-12-14 20:34:53 originally posted:

Hi Yihui,

If I save Rnw file in a different directory from output files, such as .tex, .pdf, etc, and when I run R(knitr) in the output directory, the dep_auto fails to find cache files.

For example, a project is created in the ./projectK, and the Rnw file is saved in ./projectK/src/test.Rnw. Now R and knitr are running from ./projectK/analysis using knit("../src/test.Rnw") and the root.dir='./projectK/analysis'. The following warning shows up:

Warning: file cache/__objects not found

And the dep_auto fails. Could you please help on this issue?

Thanks,

Hui Yao

Guest *Hyao* @ 2013-01-15 01:20:14 originally posted:

Hi Yihui,

I found a situation where the cache seems not work well. I hope that you could help me on it. The sample code is listed as below.

In the testCache.Rnw:

<<setup, cache="FALSE">>=

global chunk options

opts_chunk$set(cache=T, autodep=T,cache.path='cache/test-knitr-cache-KB-03-')
dep_auto() # figure out dependencies automatically
@

<<a,cache=f>>=
source("../src/test-knitr-cache-KB-03.R")
x=testCache(5)
c=5
@

<>=
(y = x + 2)
@

In addition, in the "../src/test-knitr-cache-KB-03.R":

testCache <- function(x=100){
2*x+50
}

In this case, if I change the function, testCache, the x can be updated and y CANNOT be updated accordingly, because y is cached. However, if the first chunk gets cached as, <<a,cache=t>>, even x CANNOT be updated, since literally nothing is changed in testCache.Rnw.

I know that the function, testCache, could be copied and pasted into the testCache.Rnw to solve the problem. But in some situations, I hope that the report is simple and clean, and thus would like to save utility functions in to a separate file. So do you know any way to make it work with cache being ON.

Thanks in advance,

Hui

Guest *Aleksey* @ 2013-10-10 19:57:04 originally posted:

Hi Yihui,

Many thanks for the package, it's a great thing.

I am having problems with caching - I cannot understand why I get cached output while cache=F in a chunk. Is my understanding correct that if a chunk uses cache=F, then all caching is off, and the chunk should be reevaluated every time the document is knit'ted?

My problem is, the simple 2-chunk example below works fine: 'xx' obtained via sample() is updated every time the document is run. However, caching in the 'test' chunk does not work as expected in my real document with 30 or so chunks. Actually, it works if the 'test' chunk is somewhere in the beginning of the document, but if I place it towards the end, then it does not: I always get the same numbers in 'xx' no matter how many times I rerun the document. Needless to say, big documents is where you really need caching to work correctly.

Curiously, even editing the 'test' chuck, e.g. adding xx+1 does not change xx, while the change in the chunk is echoed correctly. The only 2 things that seem to force caching off is (a) deleting cache, and (b) changing the line that produces xx, e.g. xx = sample(1:10, 4).

Am I missing something? Is there a way to troubleshoot this behavior?

<<setup, echo="F," cache="FALSE">>=
cache = TRUE # chuck cache

Default knitr chunk options

opts_chunk$set(background='gray90', cache=cache, fig.align='center', prompt=FALSE, fig.path='__figures/', cache.path='__cache/', tidy=TRUE, echo=TRUE, tidy.opts=list(width.cutoff=95, keep.blank.line=TRUE), size='small')
@

<<test, cache="F">>=
xx = sample(1:10, 3)
xx
@

On a side note, is there a difference in cache=F and cache=FALSE? There does not appear to be a difference in behavior, just looking at the knitr's progress I see that FALSE is interpreted as logical:

15:06:04.346: label: test (with options)
15:06:04.349: List of 1
15:06:04.352: $ cache: logi FALSE

while F as a symbol:

15:16:53.563: label: test (with options)
15:16:53.566: List of 1
15:16:53.569: $ cache: symbol F

This does not seem to change the caching behavior, just curious - does it matter for anything?

Many thanks
Aleksey

Guest *giovanni* @ 2014-06-29 12:10:19 originally posted:

Hi Yihui,

your package is great, i wish i had been less lazy and looked up sweave and knitr before.

Anyway, I wanted to inform you about a problem with the cache that gave me a headache. I use Rstudio 0.98.501 (the latest), knitr 1.6 (the latest) and R3.02 (not the latest, but I have a reason for this). I stupidly set cache = TRUE in some chunks. But in spite of all i tried one chunk kept not being evaluated.

The chunk is the following

<<pca, echo="TRUE," tidy="TRUE," fig.show="asis" ,="" out.width="10cm" ,="" out.height="10cm" ,="" fig.align="center" ,="" cache="FALSE," fig.cap="Screeplot of the baseball data" ,="" fig.pos="H">>=
The R code is irrelevant, I guess.

I tried many things (beside setting cache=FALSE), deleted the cache directory, changed several options (including setting tidy=FALSE, as you suggest) but nothing, the old plot kept appearing in the document. The fantastic thing is that the plot file kept reappearing after I wiped all the files in the figure directory. I have not set evaluate=FALSE globally.

The only thing that worked (I know that now it sounds obvious) was to rename the chunk to a different name. I am surprised that nobody else mentioned this problem,or am I doing something wrong? If not, it seems quite an annoying bug.At least let me know where the hidden cache is? :-)

Thanks for all your work and dedication.

Guest *Dean Attali* @ 2015-02-04 11:21:34 originally posted:

Hi Yihui, is there a way to make a cached chunk depend on an object, but not on the whole chunk from which the object was created?

Guest *Lucas Gallindo* @ 2016-06-16 14:13:19 originally posted:

I'm clueless about caching on R or knitr. I have a very simple piece of code:
<>=

Echoing the random seed and RNG for reproducibility

.Random.seed
RNGkind()
@

That fails with an ".Random.seed" object not found message. What I'm doing wrong?

Guest *Don Klinkenberg* @ 2017-07-21 14:04:49 originally posted:

Hi Yihui
I use cache as in an Rmd document, but it seems as if the code is re-evaluated every time, also when I did not change anything. I conclude that from the fact that the name of the cached file changes.
Don

This is the complete Rmd document:

knitr::opts_chunk$set(cache = TRUE, autodep = TRUE)  

I start with a simple example:

testdata <- rep(0,100)  
print(testdata)  

Guest *John Heilbron* @ 2018-05-08 00:38:48 originally posted:

Yihui,

Thanks for the package. I've made several child files in Rmd and have run those and stored contents in a cache so that the pdfs can be reproduced at the touch of a button. I want to read these child files into a parent file (and I can do that) but it seems that the parent file executes the child files and creates a separate cache of its own. I'm wondering if there's some way to direct the parent file to use the cache of the child files so that it doesn't take so long to execute. (Some of the child files took a long time to execute the first time I ran them and I'd rather not have to deal with that all over again.)

Thanks,

John

Thank you very much for your awesome work on this package!

Does caching feature also work with language engines other than R?

Thanks,
Petr