[ https://issues.apache.org/jira/browse/ARROW-6003?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16931757#comment-16931757 ]
Neal Richardson commented on ARROW-6003: ---------------------------------------- This is still the behavior in R. Slightly lower-level repro: {code:r} library(arrow) tf <- tempfile() # `iris` is a built-in dataset write.table(iris, tf, sep = ",", row.names = FALSE, col.names = FALSE) reader <- CsvTableReader$create(tf, read_options=CsvReadOptions$create(column_names="just one name")) reader$Read() ## Error in csv___TableReader__Read(self) : Invalid: Empty CSV file {code} That function is defined as {code:cpp} std::shared_ptr<arrow::Table> csv___TableReader__Read( const std::shared_ptr<arrow::csv::TableReader>& table_reader) { std::shared_ptr<arrow::Table> table; STOP_IF_NOT_OK(table_reader->Read(&table)); return table; } {code} And {{STOP_IF_NOT_OK}} is {code:cpp} #define STOP_IF_NOT(TEST, MSG) \ do { \ if (!(TEST)) Rcpp::stop(MSG); \ } while (0) #define STOP_IF_NOT_OK(s) STOP_IF_NOT(s.ok(), s.ToString()) {code} so assuming that your python code is doing essentially the same thing, maybe the difference in R is that macro, i.e. the {{s.ToString()}} maybe is not including the specific error message? > [C++] Better input validation and error messaging in CSV reader > --------------------------------------------------------------- > > Key: ARROW-6003 > URL: https://issues.apache.org/jira/browse/ARROW-6003 > Project: Apache Arrow > Issue Type: Improvement > Components: C++ > Reporter: Neal Richardson > Assignee: Neal Richardson > Priority: Major > Labels: csv > > Followup to https://issues.apache.org/jira/browse/ARROW-5747. The error > message(s) are not great when you give bad input. For example, if I give too > many or too few {{column_names}}, the error I get is {{Invalid: Empty CSV > file}}. In fact, that's about the only error message I've seen from the CSV > reader, no matter what I've thrown at it. > It would be better if error messages were more specific so that I as a user > might know how to fix my bad input. -- This message was sent by Atlassian Jira (v8.3.2#803003)