Error in if (left == 0) break : argument is of length zero

[cwaldock1]

Running a number of models from qgam package (https://github.com/mfasiolo/qgam) and run into this error. Can't provide reproducible example but here is example code. The error seems to be quite random.

Thought you'd like to know, and would be good to know if it's to do with my setting up of the clusters.

#  Detect clusters
parallel::detectCores() # 4 cores

# Create groups
group <- rep(1:4, 
             length.out = nrow(TestData %>% group_by(Group1, Group2) %>% nest()))
TestData2 <- bind_cols(tibble(group), TestData %>% group_by(TaxonomicName, Method) %>% nest())
TestData2 <- TestData2 %>% unnest(data)

# Create clusters
cluster <- create_cluster(cores = 4)

# Create partition dataframe
TestDataParty <- TestData2 %>% partition(group, cluster = cluster)

# Load library for within each cluster
cluster_library(TestDataParty, "qgam")

# Run models in parallel 
TestDataPartyOutput <- TestDataParty %>% group_by(Group1, Group2) %>% 
  do(Model1 = tryCatch(qgam(y ~ s(x, k = 3), qu = 0.95, data = .), error = function(e) NA),
     Model2 = tryCatch(qgam(y ~ s(x, k = 4), qu = 0.95, data = .), error = function(e) NA),
     Model3 = tryCatch(qgam(y ~ s(x, k = 5), qu = 0.95, data = .), error = function(e) NA),
     Model4 = tryCatch(qgam(y ~ s(x, k = 7), qu = 0.95, data = .), error = function(e) NA))

[hadley]

Can you please provide a minimal reprex (reproducible example)? The goal of a reprex is to make it as easy as possible for me to recreate your problem so that I can fix it: please help me help you!

If you've never heard of a reprex before, start by reading "What is a reprex", and follow the advice further down the page.

[cwaldock1]

Have only run into the error once, and thought it might be a bug somewhere in multidplyr. The example is a similar scale to the data I was running at the time.

Cheers

## -- ## 
library(devtools)
install_github("mfasiolo/qgam")
library(qgam)
library(dplyr)
library(multidplyr)
library(purrr)

# Create tibble
# Groups
TestData <- as_data_frame(data.frame(Group1 = sort(as.factor(rep(1:3000, length.out = 6000))), 
                          Group2 = as.factor(rep(1:2))))

# Variables
TestData <- TestData %>% 
  group_by(Group1, Group2) %>% 
  do(x = seq(1:1000),
     y = rnorm(1000)) %>% 
  unnest(x,y)

# Optional subset to smaller scale for speed. 
#TestData <- TestData[1:10000,]

#  Detect clusters
parallel::detectCores() # 4 cores

# Create groups
group <- rep(1:4, 
             length.out = nrow(TestData %>% group_by(Group1, Group2) %>% nest()))
TestData2 <- bind_cols(tibble(group), TestData %>% group_by(Group1, Group2) %>% nest())
TestData2 <- TestData2 %>% unnest(data)

# Create clusters
cluster <- create_cluster(cores = 4)

# Create partition dataframe
TestDataParty <- TestData2 %>% partition(group, cluster = cluster)

# Load library for within each cluster
cluster_library(TestDataParty, "qgam")

# Run models in parallel 
TestDataPartyOutput <- TestDataParty %>% group_by(Group1, Group2) %>% 
  do(Model1 = tryCatch(qgam(y ~ s(x, k = 3), qu = 0.95, data = .), error = function(e) NA),
     Model2 = tryCatch(qgam(y ~ s(x, k = 4), qu = 0.95, data = .), error = function(e) NA),
     Model3 = tryCatch(qgam(y ~ s(x, k = 5), qu = 0.95, data = .), error = function(e) NA),
     Model4 = tryCatch(qgam(y ~ s(x, k = 7), qu = 0.95, data = .), error = function(e) NA))

[mschilli87]

I ran into the same problem and 'resolved' it by re-initiating my cluster. I did not change any code, just re-ran the cluster setup (incl. library loading and data copying).

[hadley]

Seems unlikely to be a bug in multidplyr, but if you're still interested, and can create a simple reprex, please file a new issue.