R: Histogram with large data set - r

Question

Goal
From the data.frame d, I am trying to make a histogram of the column cMPerSite weighted by bpInPiece. In other words, bpInPiece is the number of observations at each cMPerSite value.
The Y-axis should represent densities and the X-axis should be on a log scale.
Attempts
I could do something like (which could be improved by pre-allocating the memory size for x).
x = c()
for (row in 1:nrow(d))
{
x = c(x, rep(d$cMPerSite[row],d$bpInPiece[row]))
}
hist(x,breaks=100, freq=FALSE)
but this becomes completely impractical when there is too much data (I have about 10 millions rows in my full data set) because x becomes too large to be stored in the RAM. Also, putting the X-axis in log scale is, I think, necessarily a bit of a mess.
Alternatively, I would have thought I could do
ggplot(d) + geom_histogram(aes(x = cMPerSite, y=bpInPiece), stat="identity") + scale_x_log10() + theme_classic(25)
Warning: Ignoring unknown parameters: binwidth, bins, pad
but, for some reason I do not understand, nothing gets displayed. Also, I am not sure how to put the Y-axis in density rather than count.
I suppose the bin size should vary logarithmically as the X-axis varies but that's confuses me as it would result in bins gathering an "artificial" high number of observations. Not sure how histograms are typically displayed with log scale X axis. Note that ggplot(d) + geom_histogram(aes(x = cMPerSite, y=bpInPiece), stat="identity") does not display anything either so the problem is not only a question of log scale on the X-axis.
Can you help me to make this histogram?
Subset of my data
structure(list(chrom = c(1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1), end = c(241608,
612298, 715797, 956634, 983330, 1190613, 1236417, 1330208, 1391915,
1464000, 1911436, 1913462, 2092038, 2169783, 2354812, 2363639,
2544241, 2551672, 2575287, 2589721, 2659117, 2884565, 3037319,
3100967, 3152276, 4319658, 4335072, 6301896, 6550219, 6596684,
7132319, 7435267, 7469158, 7604030, 7937619, 8131876, 9359659,
9598491, 9945959, 10262757, 10392172, 10646861, 10816847, 11094415,
11360199, 11964985, 12220179, 12222166, 12389943), cMInPiece = c(0,
1e-07, 1e-07, 0.7118558, 9.99999999473644e-08, 0.9540829, 9.99999998363421e-08,
0.4967211, 1.244988, 0.2137991, 8.808171, 0.500545200000001,
1.5721302, 1.6856566, 2.2552469, 1.0000000116861e-07, 2.6973586,
0.355113100000001, 0.355233800000001, 1.0000000116861e-07, 1.4903822,
2.8174978, 1.0000000116861e-07, 0.355231, 1.0000000116861e-07,
8.2735924, 0.425817699999996, 6.4568106, 0.372779399999999, 0.363684999999997,
0.181640399999999, 0.177473599999999, 1.0000000116861e-07, 0.177463800000005,
0.355294099999995, 1.0000000116861e-07, 1.6101482, 1.0000000116861e-07,
0.533477099999999, 0.355287800000006, 9.99999940631824e-08, 1.0000000116861e-07,
1.0000000116861e-07, 1.0000000116861e-07, 1.0000000116861e-07,
1.0000000116861e-07, 9.99999940631824e-08, 1.0000000116861e-07,
1.0000000116861e-07), bpInPiece = c(241608, 370690, 103499, 240837,
26696, 207283, 45804, 93791, 61707, 72085, 447436, 2026, 178576,
77745, 185029, 8827, 180602, 7431, 23615, 14434, 69396, 225448,
152754, 63648, 51309, 1167382, 15414, 1966824, 248323, 46465,
535635, 302948, 33891, 134872, 333589, 194257, 1227783, 238832,
347468, 316798, 129415, 254689, 169986, 277568, 265784, 604786,
255194, 1987, 167777), cMPerSite = c(1e-16, 2.69767190914241e-13,
9.66192910076426e-13, 2.95575762860358e-06, 3.74587953054257e-12,
4.60280341369046e-06, 2.18321543612659e-12, 5.29604226418313e-06,
2.01757985317711e-05, 2.96593049871679e-06, 1.96858790977928e-05,
0.000247060809476802, 8.80370374518411e-06, 2.16818650717088e-05,
1.21886131363192e-05, 1.13288774406491e-11, 1.49353750235324e-05,
4.77880635176962e-05, 1.50427186110523e-05, 6.92808654348135e-12,
2.14764856764078e-05, 1.24973288740641e-05, 6.54647349127419e-13,
5.58118086978381e-06, 1.94897583598608e-12, 7.08730509807415e-06,
2.76253860127155e-05, 3.28286140498591e-06, 1.50118756619403e-06,
7.82707414182711e-06, 3.39112268615754e-07, 5.85821989252278e-07,
2.95063589650969e-12, 1.31579423453352e-06, 1.06506539484214e-06,
5.14781970114898e-13, 1.31142734506016e-06, 4.18704366117646e-13,
1.53532728193675e-06, 1.1214963478305e-06, 7.72707909154135e-13,
3.92635728942395e-13, 5.88283747888707e-13, 3.60272081683082e-13,
3.76245376578762e-13, 1.65347744770232e-13, 3.91858719496471e-13,
5.03271269092148e-11, 5.96029260080999e-13)), .Names = c("chrom",
"end", "cMInPiece", "bpInPiece", "cMPerSite"), row.names = c(NA,
-49L), class = "data.frame")

Answer 1

This might get you started
Assuming your data is too large to process in one step - the idea is to manually generate a histogram, which is essentially the number of observations per bin
1) Split your data.frame to a size that's manageable for your memory - N can be any number
N <- 10
L <- split(df, cut(seq_len(nrow(df)), breaks=N))
2) For each split
sum bpInPiece for each group - { i %>% group_by(G = floor(-log10(cMPerSite))) %>% summarise(sum=sum(bpInPiece)) }
Then aggregate all splits - %>% group_by(G) %>% summarise(sum = sum(sum))
Then plot - ggplot(...)
library(tidyverse)
counts <- map_df(L, function(i) { i %>% group_by(G = floor(-log10(cMPerSite))) %>% summarise(sum=sum(bpInPiece)) }) %>%
group_by(G) %>% summarise(sum = sum(sum)) %>%
ggplot(., aes(G, sum)) + geom_col()
counts