If you're worried about disk usage, BEDOPS does set operations on compressed files in a format called Starch, which offers a higher compression ratio than bzip2 and, as mentioned, allows direct set operations, without any extraction to an intermediate file.

With BEDOPS installed, you could create a Starch file of SNPs like so:

$ wget -qO- http://hgdownload.cse.ucsc.edu/goldenpath/hg19/database/snp147.txt.gz \
  | gunzip -c \
  | awk -v OFS="\t" '{ print $$2,$$3,($$3+1),$$5 }' \
  | sort-bed - \
  | starch - \
  > hg19.snp147.starch

(Adjust as needed for the SNPs you need, for the reference genome you're working with.)

An uncompressed raw BED file of SNPs is roughly 5.5 GB. Compressed as a Starch file, as shown above, it should be about 5-10% of that, around 250-500 MB.

Given a sorted, tab-delimited genomic regions file regions.bed like this:

chr1    158669597   158669598
chr11   72946311    72946312

Then you could do a bedmap operation with the compressed SNP annotations like this:

$ bedmap --echo --echo-map-id --delim '\t' regions.bed hg19.snp147.starch
chr1    158669597       158669598    
chr11   72946311        72946312    rs765534001

In this example chr1:158669597-158669598 does not associate with a SNP ID, while chr11:72946311-72946312 associates with SNP rs765534001.

So there are data hygiene issues with your input you should fix before using this toolkit:

1. Your regions should be half-open, zero-based. This means that the start and stop columns should not be equal. You can use awk to increment the stop field by 1, where start and stop are equal.
2. Your regions should be sorted with BEDOPS sort-bed for set operations to run quickly and correctly.
3. Your regions should start with chr and not Chr. This can be fixed with awk, as well.